CA2001774C - Method for identifying active domains and amino acid residues in polypeptides and hormone variants - Google Patents

Abstract

The invention provides methods for the systematic analysis of the structure and function of polypeptides by identifying active domains which influence the activity of the polypeptide with a target substance. Such active domains are determined by substituting selected amino acid segments of the polypeptide with an analogous polypeptide segment from an analog to the polypeptide. The analog has a different activity with the target substance as compared to the parent polypeptide. The activities of the segment-substituted polypeptide are compared to the same activity for the parent polypeptide for the target. A comparison of such activities provides an indication of the location of the active domain in the parent polypeptide. The invention also provides methods for identifying the active amino acid residues within the active domain of the parent polypeptide. The method comprises substituting a scanning amino acid for one of the amino acid residues within the active domain of the parent polypeptide and assaying the residue-substituted polypeptide so formed with a target substance. The invention further provides polypeptide variants comprising segment-substituted and residue-substituted growth hormones, prolactens and placental lactogens.

Description

METHOD FOR IDENTIFYING ACTIVE DOMAINS
~~ND AMINO ACID RESIDUES IN
POLYPEPTIDES AND HORMONE VARIhh'TS
Field of the Invention The invention is directed to methods for identifying the active ciomains and amino acid residues in polypeptides. It is also directed to hormone variants.
Background of the Irwention l0 Polypeptides, i.e., peptides and proteins, comprise a wide variety of biological molecules each having a specific amino acid sequence, structure and function.
Most polypeptides interact with specific substances to carry out t_he function of the polypeptide. Thus, enzymes, such as subtilisin, amylase, tissue plasminogen activator, etc., interact with and hydrolyze specific substrates at particular cleavage sites whereas prote:inaceous hormones such as human growth hormone', insulin and the like interact with specific receptors to regulate growth and metabolism. In other cases, the interaction is between the pc>lypepi:ide and a substance which is not the primary target of the polypeptide such as an ~c ~s~~~.~ ~~

immunogenic receptor. Many polypeptides are pluripotential in that they contain discrete regions which interacts wit',h different ligands or receptors, each of which produces a discrete biological effort.
!5 For examplEa, human growth hormone (hGH) is diabetogenic and lypogenic in adults and induces long bone growth in children.
Efforts have been made to modify the primary functional properties of naturally occurring polypeptides by modifying amino acid sequence. One approach has been to substitute one or more amino acids in the amino acid sequence of a polypeptide with a differeni~ amino acid. Thus, protein 1!5 engineering by in _vitro mutagenesis and expression of cloned genes reportedly has been applied to improve thermal or o:~cidative stability of various proteins.
Villafranca, J.E., et al. (1983) Science 222, 782-788: Perry, L.J., et al. (1984) Science 226, 555-2~~ 557; Estell, D.A., et al. (1985) J. Biol. Chem. 260, 6518-6521; Rosenberg, S., et al. (1984) Nature (London) 312,. 77-80; Courtney, M., et al. (1985) Nature LondonZ 3:~3, 149-157. In addition, such methods have reportedly been used to generate enzymes 2:5 with altered substrate specificities. Estell, D.A. , et al. (1986) Science ~, 655-663: Craik, C.S., et al. (1985) Science ~8, 291-297; Fersht, A. R. , et al. (1985) Mature jLondon) 3~4_, 235-238; Winther, J.R., et al. (1985) Carlsberg Res. Commun. ~0, 273-3~3 284; Wells, J.A., et al. (1987) groc. Natl. Acad.
~ci. 84, 1219-1223. The determination of which amino acid residue should be modified has been based primarily on tlhe crystal structure of the polypeptide, the effect of chemical modifications on 35 the function of the polypeptide and/or the interaction of the polypeptide with various ~~'~.~~i substances to ascertain the mode of action of the polypeptide. In some cases, an amino acid substitution has been deduced based on the differences in specific amino acid residues of related polypeptides, e.g. difference in the amino and sequence in substrate binding regions of subtilisins having different substrate specificities. Wells, J. A., et al. (1987) oc.
Natl. Acad. ~~ci. U,SA 84, 5767. In other cases, the amino acid sequence of a known active region of a molecule has reportedly been modified to change that sequence to that of a known active region of a second molecule. Wharton, R. P., et al. (1985) Nature 316, 601-605, and Wharton, R. P., et al. (1984) Cell 38, 1!5 361-369 (subs,titutpon of recognition helix of phage repressor with recognition helix of different repressor); Jones, P. T., et al. (1986) Nature 321, 522-525 (subs;titut~Lon of variable region of a mouse antibody for corresponding region of human myeloma 2~) protein). While this approach may provide some predictability with. regard to the properties modified by such substitutions, it is not a methodical procedure wh~.ch would confirm that all regions and residues dete:rminat:ive of a particular property are 2!5 identified. At best, empirical estimates of the energetics for t:he strengths of the molecular contacts of s;ubstit:uted residues may be ascertained.
In this manner, 'the strengths of hydrogen bonds (Fersht, A. R., et al. (1985) Nature 3~4, 235: Bryan, 30 P., et al. (1986) Proc. Natl. Acad. Sci. USA ,$~, 3743; Wells, J. A., et al. (1986) Philos. Trans. R.
Soc. London A- ~, 415), electrostatic interactions (Wells, J. A,., et al. (1987) proc. Natl. Acad. Sci.
USA 84, 1219; Cronin, C. N., et al. (1987) T~.Am.
3'S Chem. Soc. X09, 2222), and hydrophobic and steric effects (Este:ll, D. A., et al. (1986) Science X33, ~r~~3'~.~~i 659; Chen, J. T., et al. (1987) Biochemistry ~6, 4093) have been estimated for specific modified residues. These and other reports (Laskowski, M., et al. (1987) Cold SoT~ina Harbor Symp. Quant. Biol. 52, _°. 545; Wells, J'. A., et al. (1987) Proc. Natl. Acad.
Sci. USA 84, 5167; ,Jones, P. T., et al. (1986) Nature 522; Wharton, R. P., et al. (1985) Nature X16, 601) have concluded that mutagenesis of known contact residues cau::es large effects on binding whereas 1C~ mutagenesis of non-contact residues has relatively minor effect.
A second reported approach to understand the relationship (between amino acid sequence and primary function emp7.oys ~Ln vivo homologous recombination 1_°°. between related genes to produce hybrid DNA sequences encoding h~~brid polypeptides. Such hybrid polypeptides have reportedly been obtained by the homologous recombination of trp B and trp A from E.coli and Salmonella tvphimurium (Schneider, W. P., 20 et al. (1981) Proc. Natl. Acad. Sci., USA 78, 2169-2173); alpha 1 and alpha 2 leukocyte interferons (Weber, H. and Wei:~smann, C. (1983) Nuc. Acids Res.
~, 5661); the outer membrane pore proteins ompC and phoE from ~coli K-~12 (Thommassen, J., et al. (1985) 2_°°> ~MBO 4_, 1583-1587); and thermophilic alpha-amylases from ac' lus st:earothermophilus and Bacillus ~.ichiniform:Ls (C;ray, G. L., et al. (1986) J. Bacterial. X66, 635-643). Although such methods may be capable of providing useful information 30 relating to amino acid sequence and function as well as useful hybrid polypeptides, as reported in the case of the h;~brid alpha amylases, it is difficult to utilize such methods to systematically study a given polypeptide t.o dei~ermine the precise regions and 3_°°> amino acid residues in the polypeptide that are active with one of the target substances for that particular molecules. This is because the site of crossover recombination, which defines the DNA and amino acid ~:equenc:e of the hybrid, is determined !5 primarily by the DNA sequence of the genes of interest and the recombination mechanism of the host cell. Such methods do not provide for the predetermined and methodical sequential replacement of relatively small segments of DNA encoding one polypeptide with a corresponding segment from a second gene except in those fortuitous circumstances when crossover occurs near the 5' or 3' end of the gene.
The interaction of proteinaceous hormones with their 1!~ receptors ha;s reportedly been studied by several techniques. One technique uses hormone peptide fragments to map the location of the receptor binding sitess on t:he hormone. The other technique uses compet~~tion between neutralizing monoclonal 2i) antibodies and peaptide fragments to locate the receptor binding site by epitope mapping. Exemplary of these techniques is the work reported on human growth hormone (hGH:).
Human growth hormone (hGH) participates in much of 2!5 the regulation of normal human growth and development. This, 22,000 dalton pituitary hormone exhibits a multitude of biological effects including linear growth (somatogenesis), lactation, activation of macrophages, insulin-like and diabetagenic effects 3n among others.. See Chawla, R. K. (1983) Ann. Rev.
Med. ~4_, 519;. Edwards, C. K., et al. (1988) Science 239, 769: Th,orner, M. O., et al. (1988) J. Clin.
Invest. ~1,, 745. Growth hormone deficiency in children 1~eads to dwarfism which has been ~~~~~..'~ ~'4 successfully treated for more than a decade by exogenous administration of hGH. There is also interest in the antigenicity of hGH in order to distinguish among genetic and post-translationally modified forms of :hGH (Lewis, U. J. (1984) Ann. Rev.
Physiol. 46, 33) to characterize any immunological response to :hGH when it is administered clinically, and to quantify circulating levels of the hormone.
hGH is a member of a family of homologous hormones that include placental lactogens, prolactins, and other genetic' and apecies variants of growth hormone.
Nichol, C. 5~., et al. (1986) Endocrine Reviews 7, 169. hGH is unusual among these in that it exhibits broad species specificity and binds monomerically to either the cloned somatogenic (Leung, D. W., et al.
(1987) Nature 330, 537) or prolactin receptor (Boutin, J. 1M., et al. (1988) Cell 53, 69). The cloned gene for hGH has been expressed in a secreted form in EschEaricha coli (Chang, C. N., et al. (1987) Gene 55, 189) and its DNA and amino acid sequence has been reported (Goeddel, et al. (1979) Nature X81, 544: Gray, et al. (1985) Gene ~9, 247). The three-dimensional struciture of hGH is not available.
However, the threw-dimensional folding pattern for porcine growth hormone (pGH) has been reported at moderate resolution and refinement (Abdel-Meguid, S.
S., et al. (1987) Proc. Natl. Acad. Sci. USA $4_, 6434).
Peptide fragments from hGH have been used in attempts to map the location of the receptor binding site in hGH. Li, C. H. (1982) Mol. Cell. Biochem. 46, 31;
Mills, J. B., et al. (1980) Endocrinoloctv 107, 391.
In another report, a fragment consisting of residues 96-133 was isolai~ed after proteolysis of bovine ~~~?a.'~'~

growth hormone. This fragment was reported to bind to a growth hormone receptor. Yamasakin, et al.
(1970) Bioche:mistr5r 9_, 1107. However, when a larger peptide containing residues 1-133 was produced by recombinant methodology, no detectable binding activity was observed. Krivi, G. G., et al., International Symposium on Growth Hormone: Basic and Clinical Ash>ects, Abstract I-18, Final Program, sponsored by Serono Symposia, USA, June 14-18, 1987.
These results a:re clearly irreconcilable and demonstrate the potential unreliability of using peptide fragments to map receptor binding sites on a proteinaceous hormone, especially for those where the binding site is composed of two or more discontinuous and/or conformationally dependent epitopes.
The use of neutralizing monoclonal antibodies to locate the rsacepto:r binding sites by epitope mapping has similar 7.imitations. For example, a monoclonal antibody was reportedly used in a receptor binding assay to compete with the hGH receptor for a peptide consisting oo resi.dues 98-128 of hGH. Even though the peptide !38-128 of the hGH hormone only binds to the neutralizing monoclonal antibody, it was proposed that this region contains the receptor binding site based on these competitive studies. Retegin, L. A., et al. (1982) ndoc:rinology , 668.
Similar approaches. have been used in attempts to identify antigenic sites on the hGH hormone. Epitope mapping of twenty-seven different monoclonal antibodies to hGH by competitive binding reportedly resolved only four different antigenic sites on the hormone. Surowy, T. K., et al. (1984) Mol. Immunol.
~1, 345; Aston, R.. , et al. (1985) Pharmac. Ther. ~7, ~~~~r~t/14 403. This ;strategy, however, did not locate the epitopes on the amino acid sequence of the hormone.
Another approach to defining antigenic sites has been to test the binding of antibodies to short linear .°i peptides over the protein of interest. Geysen, H.
M., et al. (1984) Proc. Natl. Acad. Sci. USA 81, 3998: Geysers, H. M. (1985) Immunol. Today _6, 364.
However, this approach suffers from the same limitations of using linear peptide fragments to locate receptor binding sites. To be useful, the linear sequence must be capable of adopting the conformation found in the antigen for the antibody to recognize it. Furthermore, based upon the known size of antibody e;pitopes from X-crystallography (Sheriff, lei S., et al. (1'987) roc. Natl. Acad. Sci USA 84, 8075;
Amit, A. G., et al. (1986) Science X33, 747) it has been estimated that: virtually all antibody combining sites must be, in part, discontinuous (Barlow, D. J., et al. (1986) Nature 322, 747) and as a result linear fragments may not adequately mimic such structure.
Peptide fragments from hGH have also been studied by non-covalently combining such fragments. Thus, several investigators have reported the analysis of the combination o:f relatively large fragments of 2!i human growth hormone comprising either the natural amino acid seaquenc~e or chemically modified peptides thereof. Burstein, S., et al. (1979) J. of Endo.
bet. 48, 964 (amino terminal fragment hGH-(1-134) combined with carboxyl-terminal fragment hGH-(141-191) ) ; Li, C'.. H. , et al. (1982) Mol. Cell. Biochem.
46 31: Mills, J. B., et al. (1980) Endocrinologv X07, 391 (subtilisin-cleaved two-chain form of hGH).

~C~~~.''~'i -g-Similarly, the chemically modified fragment hGH-(1-134) and a chemically modified carboxy-terminal )°ragment from human chorionic somatomammotropin (also called placental lactogen), 5. (hCS-(141-191)), have been non-covalently combined, as have the chemica:Lly modified fragments hCS-(1-133) and hGH-(141-191). U.S. Patent 4,189,426. These investigators reported incorrectly that the determinants for binding to the hepatic growth 1C~ hormone recEaptor are in the first 134 amino-terminal residues o:f growth hormone (Burstein, et al.
(1978) Proc. Natl. Acad. Sci. USA 75, 5391-5394).
Clearly, such. techniques are subject to erroneous results. Moreover, by utilizing two large fragments 1~~ this technique is only potentially able to localize the function i:o one or the other of the two fragments used in such combinations without identification of the specific residues or regions actively involved in a particular interaction. A review of some of the 2C~ above techniques amd experiments on hGH has been published. Nichol,, C. S., et a1. (1986) Endocrine Rev. 7, 169-2()3.
An alternative approach has been reported wherein a 7 residue peptide fragment from the "deletion peptide"
2_°°> of hGH (hGH-3:2-46) 'was modified to contain amino acid residues from, analogous segments of growth hormone from other mammalian species. The effect, if any, of such substitutions, however, were not reported. See U.S. Patent 4,699,897. Nonetheless, the shortcomings 3t~ of the use of short peptide fragments are apparent since the linear sequence of such fragments must be capable of adopting the conformation found in the intact growth hormone so that it may be recognized in an in vitr~2 or ~n vivo assay.

A number of naturally occurring mutants of hGH have been identified. These include hGH-V (Seeberg, P. H.
(1982) DNA 1, 2a9; U.S. Pat. Nos. 4,446,235, 4,670,393 and 4,665,180) and 20K hGH containing a deletion of reaidue:~ 32-46 of hGH (Kostyo, J. L. , et al. (1987) Biochemica et Bioph~sica Acta X25, 314;
Lewis, U. J., et ;al. (1978) J. Biol. Chem. 253, 2679).
One investigai=or has reported the substitution of cysteine at position 165 in hGH with alanine to disrupt the <iisulf.ide bond which normally exists between Cys-53 and Cys-165. Tokunaga, T., et al.
(1985) ~ur. ,7. Bi~ochem. 153, 445. This single substitution produced a mutant that apparently retained the tertiary structure of hGH and was recognized by receptors for hGH.
Another investigator has reported the in vitro synthesis of hGFon a solid resin support. The first report by this invs~stigator disclosed an incorrect 188 amino acid sequence for hGH. Li, C. H. , et al.
(1966) J. Am. Chem. Soc. 88, 2fl50; and U.S. Pat. No.
3,853,832. A second report disclosed a 190 amino acid sequence. U.S. Pat. No. 3,853,833. This latter sequence is also incorrect. In particular, hGH has an additional glutamine after position 68, a glutamic acid rather than glutamine at position 73, an aspartic acid rather than asparagine at position 106 and an asparagine rather than aspartic acid at position 108.
In addition t.o the foregoing , hybrid interferons have been reported which nave altered binding to a particular monoclonal antibody. Camble, r. et. al.
Properties of Interferon-a2 Analogues Produced from ~~~~~~ I

Synthetic Genes in Peptides: Structure and Function, Proceedings of the Ninth American Peptide Symposium, (1985) eds. Deber et. al., Pierce Chemical Co., Chicago, I11., pp.375-384. As disclosed therein, amino acid residuea 101-114 from a-1 interferon or residues 98-114 from y-interferon were substituted into a-2 interferon. a-2 interferon binds NK-2 monoclonal antibody whereas Q-1 interferon does not.
This particu7.ar region in a-2 interferon apparently was chosen because 7 of the 27 amino acid differences between a-1 and a-2 interferon were located in this region. The hybrids so obtained reportedly had substantially reduced activity with NK-2 monoclor.~al antibody. When tested for antiviral activity, such lZybrids demonstrated antiviral activity on par with the activity of wild type a-2 interferon. ~~ubstit:utions of smaller sections within these regions were also reported. Sequential substitution of clusters of 3 to 7 alanine residues was also proposed. However, only one analogue [Ala-30,32,33] IFN-a2 is. disclosed.
Alanine substitution within a small peptide fragment of hen egg-white lysozyme and the effect of such substitutions on the stimulation of 2A11 or 3A9 cells 2'S has also been reported. Allen, P. M., et. al. (1987) at a x,713-715.
Others have ~report~ed that binding properties can be engineered by replacement of entire units of secondary stz-ucture~ units including antigen binding loops (Jones, P.T., et al. (1986) Nature ~, 522-525) or I~NA recognition helices (Wharton, R.P., et al. (1985) ature 316,601-605).

~~~'~.'7 i ~'~

The references di~,cussed above are provided solely for their di:>closure prior to the filing date of the present application, and nothing herein is to be construed as an admission that the inventors are not entitled to antedate such disclosure by virtue of prior invention or priority based on earlier filed applications.
Given the state of the art as exemplified by the above references, :it is apparent that a need exists In for a useful method for the systematic analysis of polypeptides so ass to ascertain the relationship between structure .and function. Accordingly, it is an object herein to provide such methods to identify the active domains within the polypeptide which 1!5 contribute to the functional activity of the polypeptide.
It is a further object herein to provide methods for determining the active amino acid residues which determine functional activity.
2c) A further object of the invention is to provide methods fo~~ systematically identifying the biologically .active domains in a polypeptide.
Further, it :is an object herein to provide hormone variants having desirable biological, biochemical and 2°.i immunogenic properties which are different as compared to t:he same properties of the hormone from which such variants are derived.

~C~~~.'~'~4 Still furthe~~ it is an object herein to provide hormone variants halving diminished activity with one biological function and substantial or increased activity with a second target substance.
Still further it is an object herein to provide hGH
variants having modified binding and/or biological activity with the somatogenic receptor for hGH and increased potency.
Still further it i:a an object herein to provide hGH
1() variants which retain one or more desirable biological properties but which also have decreased diabetogenic activity.
Further, it is an object herein to provide hPRL and hPL variants having an increased binding activity 1°_> with the somaitogenic receptor of hGH.
Further, it is an object herein to provide DNA
sequences, vectors and expression hosts containing such vectors for the cloning and expression of polypeptide variants including hGH variants.
2C~ Summary of the Invention In one aspect,, the invention provides methods for the systematic analysis of the structure and function of polypeptides ~y identifying unknown active domains which influen~~e the. activity of the polypeptide with 2~~ a first target substance. Such unknown active-domains in one aspect of the invention may comprise at least two discontinuous amino acid segments in the primary amino acid aequence of the polypeptide.Active domains are determined by substituting selected amino 3C~ acid segments of the polypeptide (referred to as the parent polype~ptide;~ with an analogous amino acid ~C~~.'~~4 segment from. an .analog to the polypeptide. The analog has a different activity with the target substance as compa~.~ed to the parent polypeptide. The segment-substituted polypeptides so formed are assayed to determine the activity of each of the segment-subst:.ituted polypeptides with the target substance. ;>uch activities are compared to the same activity for the parent polypeptide. Since the structurally analogous amino acid segments are obtained from an analog that has a different interaction with the target substance, a comparison of such activities provides an indication of the location of the active domain in the parent polypeptide.
The method further comprises identifying the active amino acid rcasidue;s within the active domain of the parent polypeptide. The method comprises substituting a scanning amino acid for one of the amino acid rcesidue;s within the active domain of the 2o parent polypeptide and assaying the residue-substituted polype~ptide so formed with a target substance. The activity of each of the residue-substituted ;polypeptides is compared to the same activity of t:he parent polypeptide. These steps are 2:'S repeated for different amino acids in the active domain until the active amino acid residues are identified.
In another a~;pect, the invention provides methods to identify difl:erent active domains and active amino 3i~ acid residues for different target substances. Such methods comprise repeating the foregoing methods with a second target.

14a According to one aspect of the present invention, there is provided a method for identifying at least a first unknown active domain in the amino acid sequence of a parent polypeptide, said active domain interacting with a first target, said method comprising; (a) substituting a first selected amino acid segment of said parent polypeptide with a first analogous polypeptide segment from an analog to said parent polypeptide to form a first segment-substituted polypeptide, said parent polypeptide and said analog having a different interaction with said first target; (b) contacting said first segment-substituted polypeptide with said first target to determine the interaction, if any, between said first target and said segment-substituted polypeptide; (c) repeating steps a) and b) using a second analogous polypeptide segment from an analog to said parent polypeptide to form at least a second segment-substituted polypeptide containing a different analogous polypeptide segment than said first segment-substituted polypeptide; and (d) comparing the difference, if any, between the activity relative to said first target of said parent polypeptide and said first and second segment-substituted polypeptides as an indication of the location of said first active domain in said parent polypeptide.
According to another aspect of the present invention, there is provided a method for identifying at least one active amino acid residue in a parent polypeptide, said method comprising: (a) substituting a scanning amino acid for a first amino acid residue at residue number N within said parent polypeptide to form an N-substituted polypeptide; (b) substituting a scanning amino acid for each of the amino acid residues at residue numbers N+1 and N-1 to said first residue to form respectively N+1- and N-1-substituted polypeptides;

14b (c) contacting each of said substituted polypeptides with a target to determine the interaction, if any, between said target and said substituted polypeptides; (d) comparing the difference, if any, between the activity of the parent polypeptide and said substituted polypeptides with said target;
(e) repeating steps (b) through (d) for increasing residue numbers if said activity difference between said target and said N+1 substituted polypeptide is greater than two-fold and for decreasing residue numbers if said activity difference between said target and said N-1 substituted polypeptide is greater than two-fold.

~~.'~ ~4 In accordance with the foregoing method, polypeptide variants area identified which have a different activity wit:h ones or more target substance as compared to ~~ pare:nt polypeptide. Such variants are produced based on the identification of the active domains or t:he identification of the active amino acid residue: in the active domain which determine the activity of the parent polypeptide with a target substance.
The invention further comprises growth hormone, prolactin, and placental lactogen variants comprising at least three portions. The first portion corresponds i~o at least a part of the amino acid sequence of a parent hormone, the third portion 1'S corresponds t:o the amino acid sequence of at least part of the same parent hormone, and the second portion correaponda to an amino acid sequence of an analog to they parent hormone. The second portion is analogous to those amino acid residues of the parent 2i) hormone not contained between the first and third portions of the polypeptide variant.
The invention also includes specific human growth hormone human prolactin and human placental lactogen variants comp~risinc~ segment-substituted and residue-2!5 substituted variants of hGH.
Brief Description o~f the Drawings Fig. 1 depicta the strategy used to identify active domains.
Fig. 2 shows the conserved and variable amino acid 3i) residues amongst t:he amino acid sequences of hGH, hPL, pGH and hPRL.

~C~~~."~"~4 Fig. 3 shows the putative low resolution structure of hGH and helical wheel projections viewed from the N-terminal start residue for each helix. Hydrophobic, neutral and charged residues are indicated by O, and ~ symbols, respectively.
Fig. 4 is a bar graph showing the relative reduction in binding of various segment-substituted hGH
variants to the soluble hGH receptor.
Fig. 5 depicas the analogous amino acids in the In active domains A, C and F which interact with the somatogenic hGH receptor.
Fig. 6 depict, the relative binding positions of the somatogenic receptor and eight monoclonal antibodies to hGH.
1!5 Fig. 7 is a bar graph showing the relative increase or decrease in binding to the soluble hGH somatogenic receptor for various alanine-substituted hGH
variants. The stippled bar at T175 indicates that serine rather than alanine is substituted. The 20 broken bar ai. R178 indicates that asparagine rather than alanine is substituted.
Fig. 8 depicts the DNA and amino acid sequence of the hGH gene used in the examples.
Fig. 9 depicas the construction of vector pB0475 2°i which contains a synthetic hGH gene.
Fig. 10 is t:he DN.A sequence of p80475 showing the amino acid sequence for hGH.
Fig. 11 depicts the construction of vector pJ1446.

~~~~.'~"~4 Fig. 12 is the DNA sequence for pJ1446 showing the amino acid sequence for the soluble portion of the somatogenic receptor from liver.
Figs. 13 through 20 depict the epitope binding sites on hGH for each of eight different monoclonal antibodies.
Fig. 21 shows the: active amino acids involved in binding to the s;omatogenic receptor in hGH and helical wheel projections for helices 1 and 4.
Fig. 22 shows the rat weight gain versus time for hGH
and hGH variants administered at 50 micrograms/kg/day.
Fig. 23 is a semilog plot of Kd ratio versus potency for hGH variants as~ compared to wild-type hGH.
1!5 Fig. 24. Competitive binding curves of [1251]hGH and cold hGH to the h.GH binding protein isolated from either human serum (O) or from E. coli KS330 cultures expressing the p7lasmid phGHr(1-238) (~). Bars represent standard deviations from the mean. Inset shows Scatchard plots that were derived from the competitive t>inding curves. The concentrations of the binding protein. from human serum and E. coli were 0.1 and 0.08 :nM, respectively.
Fig. 25. Structural model of hGH based on a folding 2!5 diagram for pGH determined from a 2.8 A resolution X-ray structure. 1?anel A shows a functional contour map of the hGH receptor epitope and Panel B shows that determined here for the hPRL receptor epitope.
The size of the closed circles corresponds to the magnitude of the disruptive effect for alanine substitution at these residues. The small circles represent > 2-fold disruption whenever the larger circles represent :> 10-fold disruption. The ~ in the hGH receptor epitope (Panel A) represents the position of E17~~A that causes greater than a four-fold increase in binding affinity.
Fig. 26. Plasmid diagram of pB0760 used for intracellular expreassion of hPRL in E. coli.
Fig. 27. Locatian of residues in hGH that strongly l0 modulate its binding to the hGH binding protein.
Alanine subsi~itutions (serine or asparagine in the case of T17~~ or 1178, respectively) are indicated that cause more than a 10-fold reduction (o), a 4- to 10-fold reduction ('1), or more than a 4-fold increase (~) in binding affinity. Helical wheel projections in regions of a~-helix reveal their amphipathic quality and the fact that in helix 4 the most important dei~ermin;ants are on its hydrophilic face (shaded).
2~D Fig. 28. C:ircula:r dichroic spectra in the far UV
(Panel A) or near 1;TV (Panel B) of hGH (-) , wild-type hPRL (--) and hPRL variant D (----) (see Table XXIII).
Fig. 29. Sequence comparison of hGH and hPRL in 2!5 regions defined by homolog and alanine scanning mutagenesis t:o be important for binding. Identical residues are shaded, and the numbering is based on the hGH sequence. Residues are circled that when mutated cause more than a 4-fold change in binding affinity.
30 Asterisks above residues indicate sites at which mutations cause a 2- to 4-fold reduction in binding affinity.

Detailed Description of the Invention In one embodiment:, the method of the invention provides for the systematic analysis of a parent polypeptide, such as human growth hormone or human prolactin, to determine one or more active domains in the polypeptide that are involved in the interaction of the parent polypeptide with a target substance.
To employ the method of the invention, one or more analogs to t:he polypeptide of interest must exist 1o which exhibitor a different activity with the target substance of intereat.
Accordingly, as used herein, "parent polypeptide"
refers to any polypeptide for which an "analog"
exists that has a different activity with a target substance as compared to the same activity for the parent polyps:ptide. Examples of such polypeptides, analogs and target substances are shown in Table I.

:~4~~.'~'~4 TABLE I
Parent Target or Assay Polypeptide _ Analog- Containing Tareet Human growth Human placenta Receptors for somatogenic, hormone lactogen, human lactogenic, diabetagenic, prolactin and lipolytic, nitrogen porcine growth retention, macrophage hormone activation and insulin-like effects of hGH; rat tibia assay, rat weight gain assay, insulin resistance assay in OB/OB mice or dog, receptors on human liver, adipose, lymphocytes, thymocytes and ovary tissue hPRL pGH Binding to human prolactin receptor Rabbit GH H~.xman GH Binding to rabbit GH

receptor receptor a-interferon Related human Binding to al interferon interferons and receptor animal interferons human tissue h~unan TGIF-~2 Human hemopoietic cell growth factor o~_- inhib:ins growth modulation (TGF-~Sl ) Epidermal growthT(~F-a Carotinocyte proliferation factor (EGF) Mouse Tissue Hiunan Tissue Mouse TNF receptor Necrosis Necrosis activity Factor (mTNF) Factor (hTNF) human granulocytemouse granulocyteGrowth and differentiation macrophage colonymacrophaF;e colonyof human bone marrow stem stimulating stimulating cells factor (hGMCSF)factor (hGMCSF) human CD-4 mouse CD-4 gp-120 from HIV virus receptor receptor Subtilisin Subtilisin succinyl-ala-ala-pro-glu-~acillus yc 1 s P-Nitroanilyd Amylilguifaciens li.cheniformis ~~~.~.~'~L'~

TABLE I
(continued) Parent Target or Assay Polvpeptide _ Anal~~g- Containing Tareet human Related human Activation of human y-interferon interferons and interferon receptor animal interferons, e.g., from mouse Insulin growth Insulin IGF-1 receptor growth factor (IGF-1) growth modulation receptor Tissue Trypsin urokinase Plasminogen (cleavage) Plasminogen fibrin (binding) Activator (tPA) The parent polypeptides, analogs and target substances in Table I, of course, are exemplary only.
Parent polypeptides also include proteinaceous material comprising one or more subunits, e_.g.
!5 succinyl coenzyme A synthetase, mitochondrial ATPase, aminoacyl tRNA synthetase, glutaine synthetase, glyceraldeh,~de-3~-phosphate dehydrogenase and aspartate transcarbamolase (see, Huang, et al.
(1982), Ann. Rev. Biochem, ,~, 935-971). In such lt) multi-subunit parent polypeptides, active domains may span the two or more subunits of the parent polypeptide. Accordingly, the methods as described in more detail hereinafter can be used to probe each of the subu:nits of a particular polypeptide to 1..°i ascertain the' active domain and active amino acid residues for a particular target which may be partially con.tainect on one subunit and partially on one or more other subunits.
The parental polype~ptide and analog typically belong 2t) to a family of polypeptides which have related functions. Moreover, such parental polypeptides and analogs ordinarily will have some amino acid sequence identity, i.e., conserved residues. Such C~ ~~.'~'~~

sequence homology may be as high as 90% but may range as low as about 15~~ to 20%.
In addition to primary sequence homology, an analog to a parent :polypeptide may be defined by the three-s dimensional frame work of the polypeptide and analog. Thus, an analog may be divergent from a parent polypeptide in amino acid sequence but structurally homologous to the parent polypeptide based on a comparison of all, or part, of the tertiary structure of the molecules. Chothia, C., et al. (1986) Embo. J.: 5, 823.
In general, terti<iry analogs can be identified if the three-dimensional structure of a possible analog is known together with that of the parent polypeptide. By performing a root means squared differences (RMS) analysis of the a-carbon coordinates, (e. g., Sutcliffe, M. J., et al. (1987) Protein Engineering 1_, 377-384), the superposition of regions having tertiary analog y, if any, are identified. If the a-carbon coordinates overlap or are within about 2.~ to about 3.5A RMS for preferably about 60% or more ~of the sequence of the test analog relative to 'the a-carbon coordinates for the parent polypeptide, the toast analog is a tertiary analog to the parent polype:ptide. This, of course, would exclude any insertions or deletions which may exist between the two sequences.
Although the above parent polypeptide and analogs disclose naturally occurring molecules, it is to be understood that parent polypeptides and analogs may comprise variants of such sequences including naturally occurring variants and variations in such sequences introduced by in vitro recombinant ~C~t~~.'~"'.~4 methods. Variants used as parent polypeptides or analogs thus may comprise variants containing the substitution, insertion and/or deletion of one or more amino acid residues in the parent polypeptide or analog. Suclz variants may be used in practicing the methods of the invention to identify active domains and/or amino acids or to prepare the polypeptide variants of the invention. Thus, the naturally occurring variant, of hGH or the recombinantly produced var~~ant containing the substitution of Cys-165 with Ala may be used as parent polypeptide or an analog so long as, they have same activity with a target. Such naturally occurring and recombinantly produced variants may contain different amino acid residues which are equivalent to specific residues in another parent po:lypeptide. Such different amino acids are equivalent if such residues are structurally analogous by way of primary sequence or tertiary structure or if they are functionally equivalent.
Further, it should be apparent that many of the parent polype:ptides and analogs can exchange roles.
Thus, non-hu:man growth hormones and their related family of analogs each can be used as a parent polypeptide amd homolog to probe for active domains.
Further, targets such as the CD-4 receptor for the HIV virus, may be used as a parent polypeptide with analog CD-4 receptors to identify active domains and amino acids responsible for binding HIV and to make CD-4 variants.
As used herein, a "target" is a substance which interacts with <i parent polypeptide. Targets include receptor:~ for proteinaceous hormones, substrates for enzymes, hormones for proteinaceous ~(~~~.'~'~4 receptors, g~anerally any ligand for a proteinaceous binding protein and immune systems which may be exposed to t:he polypeptides. Examples of hormone receptors include the somatogenic and lactogenic receptors for hGH and the receptor for hPRL. Other targets include antibodies, inhibitors of proteases, hormones that bind to proteinaceous receptors and fibrin which binds to tissue plasminogen activators (t-PA).
Generally, targets interact with parent polypeptides by contacting an "active domain" on the parent polypeptide. Such active domains are typically on the surface of the polypeptide or are brought to the surface of t:he pol.ypeptide by way of conformational change in tertiary structure. The surface of a polypeptide i.s def:ined in terms of the native folded form of the polype~ptide which exists under relevant physiological conditions, i.e. in vivo or under similar conditions when expressed in vitro. The amino acid segments and amino acid residues may be ascertained in aeveral ways. If the three dimensional crystal. structure is known to sufficient resolution, i:he amino acid residues comprising the surface of the polypeptide are those which are "surface accessible". Such surface accessible residues include those which contact a theoretical water molecule "rol.led" over the surface of the three dimensional structure.

The active domain on the surface of the polypeptide may comprise a single discrete segment of the primary amino acid a~equence of the polypeptide. In many instances, however, the active domain of a native folded form of a polypeptide comprises two or more discontinuous. amino acid segments in the primary amino acid sequence of the parent polypeptide. For example, the active domain for human growth hormone with the soma.togenic receptor is shown in Fig. 5. As indicated, domain A, C and F of the active domain are each located on discontinuous amino acid segments of the hGH mols~cule. These amino acid segments are identified i;n Fig. 4 by the letters A, C and F.
Discontinuous amino acid segments which form an active domain are ~;eparated by a number of amino acid residues which are not significantly involved in the interaction between the active domain and the target.
Typically, the separation between discontinuous amino acid segment: is usually at least about five amino acids.
The methods of the invention are directed to the detection of unknown active domains in the amino acid sequence of ~~ parent polypeptide. Except for those few cases where a three dimensional crystal structure 2:5 of a polypept:ide with its target are available, _e.g.
the crystal :structure of enzymes with inhibitors or transition state analogs, most active domains for a vast array of polyp~eptides remain unknown.
As used herein an "analogous polypeptide segment" or 3n "analogous segment"' refers to an amino acid sequence in an analog which is substituted for the corresponding sequence in a parent polypeptide to form a "segme:nt sulbstituted polypeptide" . Analogous segments typically have a sequence which results in the substitui~ion, insertion or deletion of one or more different amino acid residues in the parent polypeptide while maintaining the relative amino acid sequence of the other residues in the selected segment substituted in the parent. In general, lid analogous se<~nents are identified by aligning the overall amino acid sequence of the parent polypeptide and analog i:.o ma:~cimize sequence identity between them. Analogous segments based on this sequence alignment are chosen for substitution into the 1!5 corresponding sequence of the parent polypeptide.
Similarly, analogous segments from analogs showing tertiary homology can be deduced from those regions showing structural homology. Such analogous segments are substituted for the corresponding sequences in 2p the parent. In addition, other regions in such tertiary ho:mologa, e.g., regions flanking the structurally analogous region, may be used as analogous segments.
The analogous segment should be selected, if 2~i possible, to avoid the introduction of destabilizing amino acid :residues into the segment-substituted polypeptide. Such substitutions include those which introduce bulkier side chains, hydrophilic side chains in hydrophobic core regions.
3() Typically, the amino acid sequence of the parent polypeptide a.nd analog is known and in some cases three-dimensional crystal structures may be available. 19.n alignment of the amino acid sequence of the parent: polypeptide with one or more analogs Ro~~~.

readily reveals conserved amino acid residues in the sequences which should not be altered, at least in the preliminary analysis. Sequence alignment will also reveal :regions of sequence variation which may include the substitution, insertion or deletion of one or more amino acid residues. Those regions containing such variations determine which segments in the parent may be substituted with an analogous segment. The sub:~titution of an analogous segment from an analog may therefore result not only in the substitution of amino acid residues but also in the insertion and/or deletion of amino acid residues.
If three-dimensional structural information is available, it is possible to identify regions in the 1!~ parent polype~ptide which should not be subjected to substitution with an analogous segment. Thus, for example, the identification of a tightly packed region in a hydrophobic face of an amphiphilic helix in the parent= polypeptide should not be substituted 2~) with an analogous segment. Residues identified as such should :be regained in the polypeptide variant and only surface rEesidues substituted with analogous residues.
Generally, analogous segments are 3 to 30 amino acid 2!5 residues in length, preferably about 3 to 15 and most preferably about 10 to 15 amino acid residues in length. In Nome instances, the preferred length of the analogous segment may be attenuated because of the insertion and/or deletion of one or more amino 30 acid residues in the analogous segment as compared to the homolog or parent polypeptide. If a three dimensional structure is unavailable for the parent polypeptide, it is generally necessary to form segment substituted polypeptides with analogous ~E~~~.'~ :4 segments covering most, if not all, of the parent polypeptide. Segment-substitution of the entire amino acid sequence, however, is not always necessary. For example, fortuitous segment-s substitution:: covering only a portion of the total amino acid sequence may provide sufficient information to identify the active domain for a particular target. Thus, for example, the segment-substitution of about 15% of the amino acid sequence l0 of the parent polypeptide may provide sufficient indication oi' the active domain. In most instances, however, substantially more than about 15% of the amino acid sequence will need to be segment-substituted to ascertain the active domain.
15 Generally, about 50%, preferably about 60%, more preferably about 75% and most preferably 100% of the amino acid se:quencea will be segment-substituted if no structural ir,~format:ion is available.
As used herein, "analogous amino acid residue" or 20 "analogous reaidue'" refers to an amino acid residue in an analogous segment which is different from the corresponding' amino acid residue in the corresponding segment of a parent polypeptide. Thus, if the substitution of an analogous segment results in the 25 substitution of one amino acid, that amino acid residue is an. analogous residue.
Once the parent polypeptide and one or more analogs are identified, the analogous segments from one or more of the analogs are substituted for selected 30 segments in the parent polypeptide to produce a plurality of segment-substituted polypeptides. Such substitution is conveniently performed using recombinant DNA technology. In general, the DNA
sequence encoding the parent polypeptide is cloned ~~~t. ~~I

and manipulated so that it may be expressed in a convenient host. DNA encoding parent polypeptides can be obtained from a genomic library, from cDNA
derived from mRNA from cells expressing the parent polypeptide or by synthetically constructing the DNA
sequence (Maniatis, T., et al. (1982) in Molecular Cloning, Colds Springs Harbor Laboratory, N.Y.).
The parent DI~1A is then inserted into an appropriate plasmid or v~actor which is used to transform a host cell. Prokaryotes are preferred for cloning and expressing DNA sequences to produce parent polypeptides, segment substituted polypeptides, residue-substituted polypeptides and polypeptide variants. For example, ~ coli K12 strain 294 (ATCC
No. 31446) m,3y be used as ~ coli B, E. coli X1776 (ATCC No. 31'537), and E. coli c600 and c600hf1, E.
coli W3110 (F-, °y._, prototrophic, ATCC No. 27325), bacilli such as Bacillus subtilis, and other enterobacteriaceae such as Salmonella typhimurium or Serratia marc:esans,, and various pseudomonas species.
The preferred prokaryote is ~ coli W3110 (ATCC
27325). lahen expressed in prokaryotes the polypeptide;s typically contain an N-terminal methionine or a formyl methionine, and are not 2:5 glycosilated. These examples are, of course, intended to be illustrative rather than limiting.
In addition to prokaryotes, eukaryotic organisms, such as yeast cultures, or cells derived from multicellular organism may be used. In principle, any such cell culture is workable. However, interest has been greatest in vertebrate cells, and propagation of vertebrate cells in culture (tissue culture) has become a repeatable procedure (Tissue Culture, Academic: Press, Kruse and Patterson, ~~~~.~~I

editors (197.:)). Examples of such useful host cell lines are VERO and HeLa cells, Chinese Hamster Ovary (CHO) cell 7Lines, W138, BHK, COS-7 and MDCK cell lines.
In general, plasmid vectors containing replication and control :sequences which are derived from species compatible with the host cell are used in connection with these hosts. The vector ordinarily carries a replication :site, as well as sequences which encode proteins that are capable of providing phenotypic selection in transformed cells. For example, ~ coli may be transformed. using pBR322, a plasmid derived from an E. coli species (Mandel, M. et al. (1970) J.
Mol. Biol. 53, 154). Plasmid pBR322 contains genes 1.5 for ampicillin and tetracycline resistance and thus provides easy means for selection. A preferred vector is p130475. See Example 1. This vector contains origins of replication for phage and E. coli which allow it to be shuttled between such hosts 2~3 thereby facilitating mutagenesis and expression.
"Expression vector" refers to DNA construct containing a DNA sequence which is operably linked to a suitable cc>ntrol sequence capable of effecting the expression o1' said DNA in a suitable host. Such 2!5 control sequences include a promoter to effect transcription, an optional operator sequence to control such transcription, a sequence encoding suitable mRNp, ribosome binding sites, and sequences which control termination of transcription and 3i~ translation. The vector may be a plasmid, a phage particle, or simply a potential genomic insert. Once transformed :into a suitable host, the vector may replicate and function independently of the host genome, or ma.y, in some instances,integrate into the genome itself. In the present specification, "plasmid" ~~nd "'vector" are sometimes used interchangeably as the plasmid is the most commonly used form of vector at present. However, the !5 invention is intended to include such other forms of expression vectors which serve equivalent functions and which are, or become, known in the art.
"Operably linked" when describing the relationship between two I)NA or polypeptide regions simply means li) that they are functionally related to each other.
For example, a pre~sequence is operably linked to a peptide if it functions as a signal sequence, participating in tree secretion of the mature form of the protein roost probably involving cleavage of the 1..°i signal sequence. A promoter is operably linked to a coding sequence if it controls the transcription of the sequences a ribosome binding site is operably linked to a coding sequence if it is positioned so as to permit translation.
2U Once the parent polypeptide is cloned, site specific mutagenesis (Cartel.-, P. , et al. (1986) ~ ~Jucl. Acids Res. 13, 4331; Zoller, M. J., et al. (1982) ucl.
Acids Res. ~, 6487), cassette mutagenesis (Wells, J.
A., et al. (1985) Gene 34, 315), restriction 2..'i selection mut:agenesis (Wells, J. A., et al. (1986) Philos. Tran:~. R. Soc. London SerA ~7, 415) or other known t~_chniques may be performed on the cloned parent DNA to F>roduce "segment-substituted DNA
sequences'' which encode for the changes in amino acid 3U sequence defined by the analogous segment being substituted. When operably linked to an appropriate expression vector, segment-substituted polypeptides are obtained. In some cases, recovery of the parent polypeptide or segment-modified polypeptide may be ~C~~~.'7"~4 facilitated by expressing and secreting such molecules from the expression host by use of an appropriate :signal sequence operably linked to the DNA sequence encoding the parent polypeptide or !5 segment-modified po~lypeptide. Such methods are well-known to those skilled in the art. Of course, other methods may be employed to produce such polypeptides and segment-s;ubstit:uted polypeptides such as the l~n vitro chemical synthesis of the desired polypeptide li) (Barany, G., et al. (1979) in The Peptides (eds. E.
Gross and J. Meien:hofer) Acad. Press, N.Y., Vol. 2, pp. 3-254).
Once the different segment-substituted polypeptides are produced, they are contacted with a target for 1!p the parent polypeptide and the interaction, if any, of the target and each of the segment-substituted polypeptides is determined. These activities are compared to 'the activity of the parent polypeptide with the same target. If the analog has a 20 substantially altered activity with the target as compared to 'the parent polypeptide, those segment substituted polypeptides which have altered activity with the target presumptively contain analogous segments whiclh define the active domain in the parent 2 °.i polypeptide .
If the analog has an activity with the target which is greater than that of the parent polypeptide, one or more of tlhe segment-substituted polypeptides may demonstrate a~n increased activity with that target 30 substance. Such a result would, in effect, identify an active domain in the analog and an appropriate region in the parent polypeptide which can be modified to enhance its activity with that target substance. .Such an event is most likely when the ~~~~.'~'~4 region in the analog responsible for the target interaction is contained primarily within one continuous amino acid segment. If the "active domains" of the analog comprise discontinuous regions in the amino acid sequence of the analog, enhanced activii:y in the segment-substituted polypeptide is les:> likely since the demonstration of such enhanced activity may require the simultaneous introduction of al.l active domains from the analog l0 into the segment-substituted polypeptide.
Accordingly, it is preferred that the analog have an activity with the 'target which is less than that for the parent polypeptide. In this manner, a loss in activity is observed in the segment-substituted polypeptide. However, once the active domains in a parent polypeptide are determined, that polypeptide may be used as an analog to sequentially or simultaneously substitute such active domains into a second parent: poly;peptide which lacks activity with 2~) the target for the first parent polypeptide.
Active domains in polypeptides are identified by comparing the activity of the segment-substituted polypeptide with a target with the activity of the parent polypeptide. Any number of analytical 2!~ measurements may be used but a convenient one for non-catalytic binding of target is the dissociation constant Kd of the complex formed between the segment-substituted polypeptide and target as compared to i:he Kd. for the parent. An increase or 30 decrease in x;d of about 1.5 and preferably about 2.0 per analogous residue-substituted by the segment-substitution indicates that the segment substituted is an active domain in the interaction of the parent polypeptide with the target.

'~~t~~.'~'~4 In the case o~f catalytic interaction with a target, a suitable par<imeter to measure activity relative to the parent enzyme: is to compare the ratios of kcat/Km. An increase or decrease in kcat/Km relative to the parent: enzyme of about 1.5 and preferably 2.0 per analogou:a residue-substituted indicates that an active domain has been substituted.
As used here:Ln, a "scanning amino acid" is an amino acid used to identify active amino acids within a In parent polypept.ide. A "residue-substituted polypeptide" is a polypeptide variant containing at least a sing7_e substitution of an amino acid in the parent polype~ptide with a scanning amino acid. A
"residue-substituted DNA sequence" encodes a residue 1!5 substituted polypeptide. Such DNA and polypeptide sequences may be prepared as described for the preparation of segment-substituted DNA and polypeptides.
The "active amino acid residue" identified by the 2t) amino acid scan is. typically one that contacts the target directly. However, active amino acids may also indirectly contact the target through salt bridges formed with other residues or small molecules such as H20 or ionic species such as Na+, Ca+2, Mg+2 2!i or Zn+2.
In some c2ises, the scanning amino acid is substituted f~~r an amino acid identified in an active domain of the parent polypeptide. Typically, a plurality of residue-substituted polypeptides are 30 prepared, each containing the single substitution of the scanning amino acid at a different amino acid residue within the active domain. The activities of ~d~~~.'~'~4 such residue-substituted polypeptides with a particular target substance are compared to the activity of the parent polypeptide to determine which of the amina acid residues in the active domain are involved in the interaction with the target substance. The scanning amino acid used in such an analysis may be any different amino acid from that substituted, i.e., any of the 19 other naturally occurring amino acids.
TABLE II
Isosteric Polypeptide Scanning Amino Acid Amino Acid Ala Ser, Gly Glu Gln, Asp Gln Asn, Glu Asp Asn, Glu Asn Aln, Asp Leu Met, Ile Gly Pro, Ala Lys Met, Arg Ser Thr, Ala Val Ile, Thr Arg Lys, Met, Asn Thr Ser, Val Pro Gly Ile Met, Leu, Val Met Ile, Leu Phe Tyr Tyr Phe Cys Ser, Ala Trp Phe His Asn, Gln ~(~~~."'~ ~ 4 This table u:>es the following symbols for each amino acid:
Amino Acid or residue 3-letter 1-letter thereof - symbol s~rmbol Alanine Ala A

Glutamate: Glu E

Glutamine: Gln Q

Aspartate: Asp D

Asparagine Asn N

Leucine Leu L

Glycine Gly G

Lysine Lys K

Serine Ser S

Valine Val V

Arginine Arg R

Threonine: Thr T

Proline Pro P

Isoleucine Ile I

Methionine Met M

Phenylalanine Phe F

Tyrosine Tyr Y

Cysteine Cys C

Tryptophan Trp W

Histidine His H

Most preferably, the scanning amino acid is the same for each residue substituted polypeptide so that the effect, if any, an the activity of the residue-substituted polypeptides can be systematically !5 attributed to the change from the naturally occurring amino acid reaidue to a uniform scanning amino acid residue.
In some case:>, the substitution of a scanning amino acid at one ~or more residues results in a residue-it) substituted polypeptide which is not expressed at levels which allow for the isolation of quantities sufficient to carry out analysis of its activity with a target. In such cases, a different scanning amino acid, preferably an isosteric amino acid, can be 1!5 used.

~~°~~..'~ r The most preferred scanning amino acids are relatively small, neutral amino acids. Such amino acids include alanine, glycine, serine and cysteine.
Alanine is t;he preferred scanning amino acid among this group because it eliminates the side-chain beyond the beta-carbon and is less likely to alter the main-chain conformation of the residue-substituted F~olypeptide. Alanine is also preferred because it is the most common amino acid. Further, it is frequently found in both buried and exposed positions (Creighton, T. E., in The Proteins (eds.
W.H. Freeman & Co., N.Y.); Chothia, C. (1976) J. Mol.
Biol. 150, 1). :If alanine substitution does not yield adequate amounts of residue-substituted lei polypeptide, an isosteric amino acid can be used.
Alternatively, the following amino acids in decreasing order of preference may be used: Ser, Asn and Leu.
The use of scanning amino acids is not limited to the identification of active amino acids in an active domain ascertained by the analysis of segment-substituted F~olypeptides. If, for example, one or more amino acids in a parent polypeptide are known or suspected to be involved in the interaction with a 2°_> target, scanning amino acid analysis may be used to probe that :residue and the amino acid residues surrounding it. Moreover, if the parent polypeptide is a small peptide, e.g., about 3 to 50 amino acid residues, s~~anning amino acid analysis may be carried out over the entire molecule.
Once the active amino acid residues are identified, isosteric amino acids may be substituted. Such isosteric substitutions need not occur in all instances and may be performed before any active ~~~~.'~~"~4 amino acid is identified. Such isosteric amino acid substitution is performed to minimize the potential disruptive effects on conformation that some substitutions can cause. Isosteric amino acids are shown in Table II.
Active amino acid residues can be identified by comparing the activity of the residue-substituted polypeptide with a target as compared to the parent.
In general, a two-fold increase or decrease in Kd indicates that the residue substituted is active in the interaction with the target. Similarly, in the case of cata7~ytic interaction with a target, a two fold increase or da_crease in kcat/Km relative to the parent enzymes indicates that an active residue has 1:5 been substituted.
When a suspecaed or known active amino acid residue is subjected to :>canning amino acid analysis the amino acid residues immediately adjacent thereto should be scanne d. Three residue-substituted 2i7 polypeptides are made. One contains a scanning amino acid, preferably a~~.anine, at position N which is -the suspected or known active amino acid. The two others contain the ::canning amino acid at position N+1 and N-1. If each substituted polypeptide causes a 2!5 greater than about two-fold effect on Kd or kcat/Km for a target, the scanning amino acid is substituted at position N+2 and N-2. This is repeated until at least one and preferably four residues are identified in each direction which have less than about a two-fold effect on Kd or kcat/Km or either of the ends of the parent polypeptide are reached. In this manner, one or more amino acids along a continuous amino acid sequence which are involved in the interaction with a particular target can be identified.
The methods of the: invention may be used to detect the active domain for more than one target of a particular parent polypeptide. Further, active amino acid residues within the different active domains may be also ident=ified by the methods herein. Once two or more active domains and active amino acid residues are identified for the different targets of a particular polypept:ide, various modifications to the parent polypeptide may be made to modify the interaction between the parent polypeptide and one or more of the targets,. For example, two active domains on the surface of hGH have been identified for the somatogenic and prolactin receptor. In this particular case, the active domains overlap.
Accordingly, there are a number of common active amino acid residues which interact with the 2'S somatogenic and prolactin receptors. Various modifications to hGH may be made based on this information as disc:ribed in more detail hereinafter.
In some instances, the active domain for different targets will not overlap. In such situations, modification of the active amino acids in the parent polypeptide for one receptor can be substituted with different amino acids to reduce or enhance the interaction of that active domain with its target, ~~~~~.'~"~4 thus shifting the physiological effect of such a variant.
As used herein, the term "modified interaction"
refers to a polype;ptide variant wherein one or more !5 active domains have been modified to change the interaction of the variant with a target as compared to the parent polypeptide. A modified interaction is defined as at: least a two-fold increase or decrease in the interaction of the polypeptide variant as compared to the interaction between the parent polypeptide a;nd a particular target.
The interaction between a target and a parent polypeptide, ~polypeptide variant, segment-substituted polypeptide ~3nd/or residue-substituted polypeptide 1.°> can be measured by any convenient in vitro or l~n vivo assay. Thus, in vitro assays may be used to determine any detectable interaction between a target and polypeptide, _e..g. between enzyme and substrate, between hormone and hormone receptor, between antibody and antigen, etc. Such detection may include the measurement of color metric changes, changes in radioactivity, changes in solubility, changes in molecular weight as measured by gel electrophoresis and/or gel exclusion methods, etc.
2_°°> In vivo assays include, but are not limited to, assays to detect physiological effects, g.g. weight gain, change in electrolyte balance, change in blood clotting time, changes in clot dissolution and the induction of antigenic response. Generally, any ,~
vivo assay may be used so long as a variable parameter exists so as to detect a change in the interaction bEatween the target and the polypeptide of interest.

R:~((3~.'7'74 ' Exemplary of the present invention is a preferred embodiment wherein the active domains and active amino acids ~of human growth hormone which determine its activity with its somatogenic receptor are identified. In carrying out this embodiment of the invention, human growth hormone variants, including segment-substituted and residue-substituted hGH
variants, ha~~e been made or identified which have different binding interactions with the somatogenic receptor for growth hormone as compared to naturally occurring human growth hormone. At least one of these human growtlh hormone variants has a higher affinity for the ;somatogenic receptor and enhanced potency for somatogenesis in rats. Others, have a 1.5 decreased activity with the somatogenic receptor.
Such hGH variants are useful as hGH agonists or antagonists and may have a higher potency for stimulating other receptors for human growth hormone since such variants will be freed from substantial 2n interaction with the somatogenic receptor. Further, such variants. are useful in immunoassays for hGH as an hGH standard or tracer. In one instance, a variant has been identified which has a significant decrease in reactivity with human and mouse serum 2!~ containing anti-hGH polyclonal antibodies. Another has the same binding affinity for the somatogenic receptor as :hGH but increased potency to stimulate growth.
The method for determining the active domains for 3n human growth hormone which interact with its somatogenic: receptor from liver is shown schematically in Fig. 1. In this approach, segments of hGH were systematically replaced with analogous sequences from analogs of hGH that are known to have 3!~ greatly reduced affinities for the cloned hGH liver r~~~~~..~~~

receptor and for monoclonal antibodies raised against hGH. Such analog:a for hGH include human placenta lactogen (hP:L), porcine growth hormone (pGH) and human prolact:in (h:PRL). These analogs have binding !5 affinities for the cloned hGH receptor that are reduced by about 100 to 10,000-fold for the somatogenic hGH receptor (hGHr) (Harrington, A. C., et al. (1986) JClin. Invest. 77, 1817; Baumann, G., et al. (1986,) JyClin. Endocrinol. Metab. 62, 137.
Such analogs .are used because homologous proteins are known to have similar three-dimensional structures even though they may have a large sequence divergence (Chothia, C. , et a:l. (1986) EMBO J. 5, 823) . In so doing, the likelihood is increased that analogous 1..°i sequence subs,titut~.ons will be readily accommodated without grossly di~~rupting the native folding of the molecule. The amino acid sequence for human growth hormone and t:he analogs hPL, pGH and hPRL are shown in Fig. 2. These latter three analogs share a sequence identity with hGH at the level of 85%, 68%
and 23%, respectively.
Referring to Fig. 1, the overall strategy is shown for identifying ons: or more active domains in human growth hormone which interact with the somatogenic 2..'i receptor for human. growth hormone (a "target" for hGH). As indicated, hGH has a positive binding activity with the target receptor, in this case, the somatogenic receptor. The hPRL, hPL and pGH
analogs, howswer, have a greatly reduced activity with that target as indicated by the minus sign.
Six segment-s;ubstit~uted growth hormones, identified by letters A through F, are formed by substituting a selected amino acid segment of hGH with an analogous amino acid segment from the hPRL analog. Each of 3°_i these selected segments are different and are chosen ~i to probe either the: entire amino acid sequence of the hGH molecule or those regions which are expected to contain the active domains. After the segment-substituted human growth hormones are prepared each is assayed against: the hGH somatogenic receptor to determine its activity. The results of such an assay as compared to hGH are indicated by + or-under the segment-modified human growth hormones in Fig. 1. A.s can be seen in Fig. 1, segment-substituted human growth hormones C and F in this schematic do not bind the somatogenic receptor.
Based on there results, those regions in the growth hormone corresponding to the analogous segments from the analog in the growth hormone variants C and F
are identified as active domains involved in the binding of hGH to i.ts somatogenic receptor.
As indicated, it is not necessary to probe the entire amino acid sequence of human growth hormone or other parental polypeptides if structural information or other data are available. Thus, low-resolution or high-resolution structural information from crystallographic studies can provide important information so as to avoid destabilizing substitutions of selected amino acid segments from a homolog. For example, the X-ray coordinates for human growth hormone are not available. However, helix wheel projections from the pGH folding model, based on the low resolution X-ray crystal structure of pGH, reveal that three of the four helices (helix 1, 3 and 4) are amphipathic with strong hydrophobic moments. SeEa Fig. 3. Eisenberg, D., et al. (1984) J. Mol. Biol,_ ~, 125. Since the hydrophobic core in polypeptides is. very tightly packed (Ponder, J.
W., et al. (1987) _J'. Mol. Biol. X93, 775), changes in suchw buried amino acid residues are generally ~~~~.""~"~4 destabilizing (Alber, T., et al. (1987) Biol. Chem.
26, 3754; Reidhaar-Olson, J. F. (1988) Science 241, 53 ) .
In addition, regions of high amino acid sequence conservation amongst members of the polypeptide family, for example the human growth hormone family, in general, need not be probed, at least initially.
This is because t:he disruption of such conserved sequences is likely to disrupt the folding of the 1~3 molecule. further, other data may suggest that certain regions of= the parent polypeptide are not involved in t:he in'~teraction with a particular target substance. For example, deletion of the N-terminal 13 amino acids of :hGH by mutagenesis (Ashkenazi, A., 1'S et al. (1987) ndocrinology ,~, 414) and a natural variant of hC~H which deletes residues 32 to 46 (the 20Kd variant; Lewis, U. J., et al. (1980) Biochem.
Bio~hys. Res" Commun. 92, 5111) have been reported not to effect: dram<~tically the binding properties to 2i~ the somatogenic receptor. In addition, the production of a two-chain derivative of hGH by limited proteolysis, which deletes some or all of the residues between 134 and 149, does not markedly effect binding to the somatogenic receptor.
2!5 Li, C. H. (1f82) ~Iol. Cell. Biochem. 46, 31~ Mills, J. B., et al. (1980) Endocrinology ~, 391.
Based on this. information, six segments of the amino acid sequence: of h.GH were selected for substitution with the corresponding analogous amino acid segments 30 from a number of analogs to hGH. These selected segments are identified as A through F in Fig. 2.
These segments are separated either by disulfide bonds, by borders of secondary structure (see Fig.
4), by areas of high sequence conservation in the growth hormone faimily or by regions previously identified as. not being involved in binding to the somatogenic receptor. Seventeen segment-substituted hGH variants were prepared which collectively .°> substituted 85 out of the 191 residues in hGH. The regions identified as A through F in Fig. 2 and the segment-substituted hGH variants prepared within each region are swzunarized in Table III.

h TA BLE II
I

Kd Region Segment- A ctual Mutagenesis (variant) probed Subs titutedSub:; titution method Kd (nM) hGH VariantIntroduced Kd(Wt) hGH None 0.34 1.0 A 11-33hPL (12-25)N12H, F25L r.s.l/ 1.4 4.1 pGH (11-33)D11A, M14V,H18Q cassette?/1.2 3.4 R19H, F25A,Q29K, hPRL (12-33)N12R, M14V,L15V,cassette 3.6 11 R16L, R19Y,F25S, D26E, Q29S,E30Q, hPRL (12-19)N12R, M14V,L15V,r.s. 5.8 17 R16L, R19Y

hPRL (22-33)Q22N, F25S,D26E,r.s. 0.29 0.85 Q29S, E30Q,E33K

B 46-52hPL (46-52)Q46H, N47D,P48S,r.s. 2.5 7.2 Q49E, L52F

pGH (48-52)P48A, T50A,S51A,r.s. 0.94 2.8 C 54-74hPL (56-64)E56D, R64M cassette 10 30 pGH (57-73)S57T, T60A,S62T,cassette 5.8 17 N63G, R64K,E65D, T67A, K70R,N72D, hPRL (54-74)F54H, S55T,E56S,cassette 23 69 I58L, P59A,S62E, N63D, R64K,E66Q, T67A, K70M,S71N, N72Q, L73K,E74D

~C~~~.'~'~4 TABLE III

(continued) Kd Region Segment- Actual Mutagenesis (variant) probed Substituted Substitution method Kd (nM) hGH Variant Introduced Kd(wt) D 88-104hPRL (88-95)E88G,Q91Y, F92H,r.s. 0.47 1.4 R94T.,S95E

hPRL (97-104;1F97R"A98G, N99M,r.s. 0.53 1.6 S100Q,L101D, V102A,Y103P, E 108-136hPL (109-112)N109I),V110D, cassette 0.61 1.8 hPRL (111-121)Y111V,L113I, cassette 0.52 1.5 K115E,D116Q, E118K,E119R, G120L,Q122E, T123G,G126L, 81277:E129S
, hPRL (126-13E~)R127I),L128V, cassette 0.58 1.7 E129Ii,D130P, G131E,S132T, P133K,R134E, F 164-190pGH (164-190)Y164S,R167K, hybrid3/ >34 >100 M170L,D171H, V173E~,F176Y, I179V,V180M, Q181K,S184R, I184F,G187S, pGH (167-181)R167K,D171H, r.s. 9.2 27 I179V,Q181K

1/ Restriction selection - Wells, J. A., et al. (1986) philos. Trans.
R. Soc. London SerA ~', 415.
Cassette mutagenesi.s - Wel.ls, J. A., et al. (1985) ene ~, 315.
3/ Recombination mutagenesis - Gray, G. L., et al. (1986) J.
~acteriol. X66, 635.

~~~.~~I

The segment-;substituted hGH variants are generally identified by the analogous segments substituted into the human growth hormone sequence . However, in some instances, not all of the analogous residues in the substituted analogous segment were maintained in a particular construction. Thus, in Table III hPL
(12-25) ident:ifies a segment-substituted hGH variant wherein amino acid: 12 through 25 of human placental lactogen (hPL) are substituted for amino acid residues 12 through 25 in the parent hGH. The effect of sux~stitui~ing this analogous segment can be determined by comparing the amino acid sequence of hGH and hPL in this region in Fig. 2. Four amino acid substitutions are generated in an hPL (12-25) variant where no other changes are made. These residues are 12, lE~, 20 and 25 for hPL (12-25).
The actual amino acid substitutions in the hPL
(12-25) variant and the other segment-substituted variants are shown in Table III. Each substitution is represented by a~ letter followed by a number which is followed by a leater. The first letter and number correspond to the amino acid at that residue number in the unmodified hGH. The last letter corresponds to the amino acid which is substituted at that posi-2.5 tion. Thus, N12H indicates that the asparagine at position 12 in hGH is substituted by histidine in the hPL (12-25) variant:.
As can be seen, some of the actual substitutions introduced do not correspond to the totality of substitutions indicated by the corresponding segments in Fig. 2. '.thus, ;hPL (12-25) would contain the four substitutions N12H, R16Q, L20A and F25L if the entire hPL (12-25) segment were substituted. The actual variant madea, however, maintained R16 and L20 ~~~~~.'~°~ .4 and therefore incorporated only two of the four substitutions, i.e., N12H and F25L, as shown in Table III. Other segment substituted variants which maintained one or more resudues of the parent hGH
include those covering regions A and E and the segment subsi:ituted variants hPL (46-52) and pGH
(167-181).
Each of the ~;egment:-substituted human growth hormone variants were assayed in an ~_n vitro system comprising displacement of ~125I~hGH from the extracellular portion of the cloned soluble hGH
receptor to quanti:Ey the relative affinities of the segment-substituted variants to the extracellular domain of the somatogenic receptor. Leung, D. W., et 1..'i al. (1987) Nature 330, 537. This truncated form of the somatoc~enic receptor exhibits the same selectivity i'or hCiH as the membrane form of the receptor (Spencer, S. A., et al. (1988) J. Biol.
Chem. 263, 7862) albeit with about a slight reduction in binding affinity (ICd = 0.3nM).
As will be described in more detail in the examples, selected segments A, C and F, comprising residues 11-19, 54-74 and 164-191, respectively, are active domains in t:he hGH molecule interactive with the 2°.i somatogenic receptor. This is based on the observed decrease in Kd of ten-fold or greater for most of the segment-substituted hGH variants containing analogous segments for hGH analogs over these regions. See Fig. 4. Of course, this does not mean that each of the amino acid residues within these active domains comprise the binding residues for the somatogenic receptor. Rather, such domains define the amino acid sequence within which such active residues can be found.

~:~~1.'7'i 4 The active domains A, C and F were further localized.
For example, the variant hPRL (12-33) was dissected into the amino and carboxy terminal variants, hPRL
(12-19) and hPRL (22-33). The results from this experiment further localized this active domain of hGH to residues 12 through 19. Similarly, the amino terminal portion of' region F (pGH (167-181)) exhibits a large redu~~tion in binding affinity. One of the most dramatic: effects was the 30-fold reduction in binding caused by hPL (56-64) which introduced only two mutations,, E561~ and R64M. Although regions A, C
and F are widely separated in the primary sequence of hGH, the tertiary folding of the hormone brings them within close proximity. See Fig. 5. These active domains form a patch that contains the amino terminus of helix 1 (active domain A), the loop from Cys-53 to the start of helix 2 (active domain C) and the central portion of helix 4 (active domain F).
In addition, eight Mabs against hGH were assayed against segment-substituted hGH variants to map the epitopes of hGH. Further, the Mab~s were used in a competitive assay with hGH and hGH variants to evaluate the ability of each of the Mabs to block the binding of the hGH receptor to hGH.
The collective results obtained from these experiments provide several lines of evidence that the substitul:ion of analogous segments into hGH do not grossly disrupt the native folding of the molecule and that the observed activity is due to a direct effect on the interaction between the somatogenic receptor and the segment-substituted hGH
variants. Firstly, the segment-substituted variants are highly s;elect:ive in disrupting binding to the ~~~~.'~'r 4 somatogenic receptor or the Mabs. Secondly, the somatogenic receptor and Mabs recognize conformation as well as s,equenc:e. The receptor and at least four of the Mabs .recognize discontinuous epitopes that are sensitive t o the protein tertiary structure.
Thirdly,, circular dichroic spectra of all of the purified variants are virtually identical to wild-type hGH. :Fourthly, all of the variants, with the exception of pGH (164-190), were expressed in essentially wild-type amounts. Resistance to proteolysis ~ v_~vo has been used as a screen for conformational integrity. Hecht, M. H., et al.
(1984) Proc. Natl. Acad. Sci. USA 81, 5685; Shortle, D. , et al . (:L985) !Genetics 110, 539.
7.5 The alteration i:n binding activity for segment-substituted hGH variants does not necessarily indicate that the substituted residues in such variant:: make direct contact with the somatogenic receptor. A disruptive mutation may not only remove a favorabls~ interaction but may introduce an unfavorable one. For example, the N12R mutation in the hPRh ( 12 -19 ) s,egment-substituted hGH variant not only changes the hydrogen bonding amide function of Asnl2, t:he Arg substitution also introduces a bulkier side chain tlhat is. positively charged. Furthermore, a number of the binding contacts may be conserved between the analogs so that not all contacts, or even regions, may be probed by generating segment-substituted hGH variants. Further, the substitution of analogous segments generates the substitution of multiple amino acid residues in the hGH molecule.
In order- to identify the specific active amino acids within the active domains A, C and F in Fig. 2, a fine structure analysis of these active domains was ~~~~.'~'~4 performed. 7Cn this analysis, residues in these three active domains were replaced sequentially with alanine. A total of 63 single Alanine mutants were made and each of their binding constants were determined for the soluble hGH receptor (shGHr) by Scatchard analysis. Leung, D. W., et al. (1988) T~.
Biol. Chem. x!63, 7862.
Based on this analysis, the amino acid residues listed in Table IV comprise residues within the hGH
molecule which are actively involved in the interaction saith the somatogenic receptor. This is based on the more than four-fold effect on the relative dissociation constant caused by the substitution. of alanine for these residues as compared to wt hGH. See Fig. 7. Preferred amino acid substitutions for these residues to form hGH
variants are shown..
TABLE IV
Preferred amino hGH Residue acid substitution R64 GEMFAQSH,KDN

Other amino acid residues which are less active with the somatogenic receptor are listed in Table V.
These residue's demonstrate generally less than two fold increases in relative Kd when substituted with 5 alanine.

~~~~.'~'~4 TABLE V

Amino acid residues in hGH showing a relative decrease in Kd when substituted with alanine (and consequently greater affinity for the somatogenic receptor) are listed in Table VI.
TABLE VI

One residuEa substituted hGH variant, E174A, surprisingly resu7Lted in a significant decrease (almost five-fold) in the dissociation constant with the somatogenic receptor. This variant, ~ in addition !5 to showing em increased binding affinity for the somatogenic receptor also exhibited an increased somatogenic potency relative to hGH in a rat weight gain assay. This and other specific residue substitutes that enhance somatogenic binding by >1.4 fold are presented in Table VII.

~:~~~.'~ ~ 4 TABLE VII
hGH variania having enhanced somatogenic binding Substituted ~H resi.dues amino acid E6.5 A

E174 A,N,Q,S,G

Other variants containing alanine substitutions not shown in Fig. 7 are listed in Table VIII.
TABLE VIII
Variant Kd (mM) ICd (var) /Kd (wt) K172A/F1'76A 201 543 N47A 0.84 2.3 Q49A 0.36 1.0 T50A 0.38 1.0 V17'.3A NE -Note NE - not expressed in shake flasks at levels which could b~: easily isolated (i.e., < - 5% of wild-type expression levels).
Once identified, the active amino acid residues for the somatogenic receptor in hGH are analyzed by substituting different amino acids for such residues other than the scanning amino acid used for the °_> preliminary analysis. The residue substituted variants in Table I:X have been made.

~~~~.'~"~4 TABLE IX
Variant Kd(nM) Kd(var)/Kd(wt) R7?V 0.44 1.3 L80D 0.78 2.3 F176Y 3.2 8.6 E174G 0.15 0.43 E174H 0.43 1.2 E174K 1.14 3.1 E174L 2.36 6.4 E174N 0.26 0.7 E174Q 0.21 0.6 E174S 0.11 0.3 E174V 0.28 0.8 R64K 0.21 0.6 D169N 3.6 10.5 Note: NE - not e:rcpressed in shake flasks at levels which could be easily isolated (i.e., < - 5% of wild-type expression le~rels).
In addition i.o the hGH variants that have been made, Table X identifies. specific amino acid residues in hGH and replacement: amino acids Which are expected to produce variants having altered biological functions.

TABLE X
wT hGH: amino _acid r~esidus~ Replacement amino acid In another embodiment, The binding epitope of hGH for the prolactin receptor was determined. hGH can bind to either the growth hormone or prolactin(PRL) receptor. As will be shown herein, these receptors compete with one another for binding to hGH
suggesting that their binding sites overlap. Scanning mutagenesis data show that the epitope of hGH for the hPRL receptor consists of determinants in the middle of helix 1 (c:ompri ing residues Phe25 and Asp26), a loop region (including I1e58 and Arg64) and the center portion of helix 4 (containing residues K168, K172, E1'74, and F176). These residues form a patch ~:~~~~.'~'~4 -5~-when mapped upon a structural model of hGH. This binding patch overlaps but is not identical to that determined for they hGH receptor as diclosed herein and by B.C. Cunnangham and J.A. Wells (1989) Science 244, 1081-1OF35. By mutating the non-overlap regions of these receptor binding sites on hGH, the preference of hGH was shifted toward the hGH receptor by >2000-fold or toward the hPRL receptor by >20-fold without loss in banding affinity for the preferred receptor. Similarly, by mutating the overlap regions it is possible to reduce binding to both receptors simultaneously by >500-fold. Such receptor selective variants of hGH should be useful molecular probes t:o link specific receptor binding events to 1.5 the various biological activities of hGH such as linear growth or lactation.
In a further embodiment, the receptor binding determinants from human growth hormone (hGH) were placed into the normally nonbinding homolog, human prolactin (hl?RL). The alanine scanning mutagenesis disclosed herein and Cunningham, B. C. & Wells, J. A.
(1989) Science 246, 1081-1085 identified important residues in hGH for modulating binding to the hGH
receptor cloned from human liver. Additional 2!5 mutations derived i°rom hPRL were introduced into hGH
to determine which hPRL substitutions within the hGH
receptor. binding site were most disruptive to binding. Thereafter, the cDNA for hPRL was cloned and expressed in Escherichia coli. It was then mutated to sequentially introduce those substitutions from hGH that. were predicted to be most critical for receptor binding. After seven iterative rounds of site-specific mutagenesis, a variant of hPRL
containing eight mutations whose association constant 3!5 was strengthened over 10,000-fold for the hGH

receptor was ideni:ified. This hPRL variant binds only six-fold weaker than wild-type hGH while sharing only 26% overall sequence identity with hGH. These results show the ;structural similarity between hGH
and hPRL, and confirm the identity of the hGH
receptor epii~ope. More generally, these studies demonstrate the feasibility to borrow receptor binding properti~as from distantly related and functionally divergent hormones that may prove useful 1~~ for the design of hybrid hormones with new properties as agonist or antagonist.
The following is presented by way of example and is not to be construed as a limitation to the scope of the invention.
1!5 Example 1 hGH Mutactenesis and Expression Vector To facilitate efficient mutagenesis, a synthetic hGH
gene was mad~a that: had 18 unique restriction sites evenly distributed without altering the hGH coding 20 sequence. The synthetic hGH DNA sequence was assembled b;y lic~ation of seven synthetic DNA
cassettes each roughly 60 base pairs (bp) long and sharing a 10 ;bp overlap with neighboring cassettes to produce the ~t05 by DNA fragment shown from NsiI to 2!i BglII. The ligated fragment was purified and excised from a polyac:rylamide gel and cloned into a similarly cut recipient vector, pB0475, which contains the alkaline: ph~~sphat:ase promoter and StII signal sequence (Chang, C. N., et al. (1987) Gene ~5, 189), 30 the origin of replication for the phage fl and pBR322 from by 1x05 through 4361 containing the plasmid origin of replication and the p lactamase gene. The sequence was confirmed by dideoxy sequence ~~~~~.''~'74 analysis (Sanger, 1?., et al. (1977) Proc. Natl. Acad.
Sci. USA, 74, 5463)"
pB0475 was construcaed as shown in Fig. 9. fI origin DNA from filamentous phage contained on a DraI, RsaI
fragment 475bp in length was cloned into the unique PvuII site o:E pBR322 to make plasmid p652. Most of the tetracycline resistance gene was then deleted by restricting p652 with NheI and NarI, filling the cohesive end:a in with DNA polymerase and dNTPs and ligating the large 3850bp fragment back upon itself to create the pla:amid po 652 . po 652 was restricted with EcoRI, EcoRV and the 3690bp fragment was ligated to a 1.300bp Eco:RI, EcoRV fragment from phGH4R
(Chang, C. N., et al. (1987) Gene 55, 189) containing the alkaline phosphatase promoter, STII signal sequence and natural hGH gene. This construction is designated a~; pB04'73. Synthetically derived DNA was cloned into pB0473 in a three-way construction. The vector pB047:3 was restricted with NsiI, BglII and ligated to <~ 240pb NsiI, HindIII fragment and a 1170bp HindII, Bc~lII fragment both derived from synthetic DNA. The resulting construction pB0475 contains DNA coding for the natural polypeptide sequence of hGH but possesses many new unique restriction sites to facilitate mutagenesis and further manipulation of the hGH gene. The entire DNA
sequence of pB0475 together with the hGH amino acid sequence is schown :in Fig. 10. The unique restriction sites in the hGH sequence in pB0475 allowed insertion 3~D of mutagenic cassettes (Wells, J. A., et al. (1985) ne ~4,, 315) containing DNA sequences encoding analogous segments from the analogs pGH, hPL and hPRL. Alternatively, the hGH sequence was modified by site specific aiutagenesis in the single stranded 3:5 pB0475 vector followed by restriction-selection ,~ 2~~ Z~74 ..

against one of the unique restriction sites (Wells, J. A., et al,. (1986) Philos. Trans. R. Soc. London SerA 317, 415).
The 17 segment-substituted hGH variants in Table III
were prepared. Each was secreted into the periplasm.ic space of E. coli at levels comparable to wild-type hGH and at levels that far exceeded the hGH-pGH hybrid described infra. The hGH and hGH
variants were expressed in E. coli W3110, tonA (ATCC
27325) grown in low phosphate minimal media (Chang, C. N., et al. (1987) Gene 55, 189).
The hGH and hGH variants were purified as follows. To 200g of cell paste four volumes (800mI) of lOmM tris pH 8.0 was added. The mixture was placed on an orbital shaker' at room temperature until the pellets were thawed. The mixture was homogenized and stirred for an hour in a cold room. The mixture was centrifuged at 7000g~ for 15 min. The supernatant was decanted and ammonium sulfate was added to 45%
saturation (277g/1) and stirred at room temperature for one hour. After' centrifugation for 30 minutes at 11,OOOg, the pellet;. was resuspended in 40m1 lOmM
tris pH 8Ø This was dialyzed against 2 liters of lOmM tri.s p::i 8.0 overnight. The sample was centrifuged or filtered over a 0.45 micron membrane.
The sample waa then loaded on a column containing 100m1 of bEAE cellulose (Fast Flow, Pharmacia, Inc.).
A gradient of from zero to 300mM NaCl IN lOmM TRIS
PH 8.0 in 8 to 10 column volumes was passed through the column. Fractions containing hGH were identified by PAGE, pooled, dialyzed against IOmM
tris H2C1 F>H 8.0 overnight. Samples were concentrated to approximately lmg/ml by Centri-PreplO~
ultrafiltration.
*Trade-Mark I

Example 2 Liomoloqous Reco inants of hGH and pGH
A random hybrid library containing various N-terminal lengths of hGH linked to the remaining C-terminal !5 portion of porcine growth hormone (pGH) was constructed by the method of random recombination of tandomly linked genes. Gray, G. L., et al. (1986) 7~.
Bacteriol. X66_, 635.
The EcoRI site of pB0475 was removed by restricting the plasmid with Ec:oRI, filling in the cohesive ends by addition of DNA polymerase and dNTPs, and ligating the plasmid back together. A new EcoRI site was then introduced just following the 3' end of the hGH gene.
This was accomplished by subcloning the 345bp BglII, EcoRV fragment of hGH-4R which contains such an EcoRI site, into similarly restricted rector from the EcoRI- pB0475 consltruction. The pGH gene (Seeburg, P. H. , et al ,~ ( 1983 ) ~ 2_, 37 ) was then introduced just downstream and adjacent to the 3' end of the hGH
gene in this construction. This was accomplished by doping an ~EcoRI, HindIII (filled in) fragment containing pC:H cDNA into the large fragment of a EcoRI, EcoRV digest of the construction described above. The resulting plasmid, pB0509, contains an 2°_i intact hGH gene with a unique EcoRI site at its 3' end followed by an intact pGH gene reading in the same direction. Due to the homology between the hGH
gene and pGH genes, a percentage of the pB0509 plasmid underwent ~ vivo recombination, to make hybrid hGH/pGH genes when transformed into ~ Soli rec+ ~I294 (.ATCC :31446). These recombinants were enriched by restricting pool DNA with EcoRI to linearize plasmids which had not undergone recombination resuT,ting in the loss of that EcoRI

~;~~~."~"~4 site. After two rounds of restriction selection and transformation into ~ oli rec+ MM294 nearly all the clones represented hybrid hGH/pGH recombinants.
Sequence analysis of 22 clones demonstrate that the hGH/pGH hybrids contained with amino terminal hGH
sequence followed ;by pGH sequence starting at amino acid residue: +19" +29, +48, +94, +105, +123 and +164.
Seven hGH-pGH hybrids having cross-over points evenly distributed over the hGH gene were obtained.
However, onl;~ the extreme carboxy terminal hybrid (hGH (1-163)-;pGH (164-191)) was secreted from E. coli at levels hi~3h enough to be purified and analyzed.
This hGH-pGH hybrid introduces three substitutions 1!5 (M170L, V173~~ and V180M) that are located on the hydrophobic face of helix 4. Accordingly, most of the sequence modifications in the helical regions A, D, E and F in Fig. 2 were designed to avoid mutations of residues an the hydrophobic face of the helices.
For example, the above hybrid hGH-pGH variant was modified to retain M170, V173, F176 and V180 because these residues are inside or boarding the hydrophobic face of helix 4.
Example 3 2-'i Expression and Purification of Soluble Human Growth Hormone Receptor from E. coli Cloned DNA sequences encoding the soluble human growth hormone receptor shGHr (Leung, D. W., et al.
(1987) Zlature ~_0, 537) were subcoloned into pB0475 to form pJ1446 (see Figs. 11 and 12).
The vector pClS.2 :>HGHR (Leung, D. W., et al. (1987) Nature 330, 537) was digested with XbaI and KpnI and the l.Okb fragment containing the secretion signal plus the 246 codon extracellular portion of the hGH
receptor- was purified (Maniatis, T. et al. (1982) in Molecular Cloning, Cold Springs Harbor Laboratory, N.Y.). This fragment was ligated into similarly cut M13-mpl8 and single-stranded DNA for the recombinant was purified (Messing, J. (1983) Methods in Enzvmolocrv, Vol. 101, p. 20) . Site-specific mutagenesis (Carte:r, P., et al. (1986) Nucleic Acids es. ~3,, 4331) wa:c carried out to introduce an NsiI
site at codon +1 using the 18 mer digonucleotide, 5'-A-AGT-GAT-~GCA-T'.CT-TCT-GG-3'. The mutant sequence was versified by dideoxy sequence analysis (Sanger, F., et al. (1977) Proc. Natl. Acad. Sci. USA 74, 5463). Double-stranded DNA for the mutant was purified and cut with NsiI and SmaI. The 900bp fragment was: isolated containing the 246 codon extracellular portion of the hGH receptor. pB0475 was cut with NsiI and EcoRV and the 4.lkb fragment (missing the synthetic hGH gene) was purified. The 900bp fragment for the receptor and the 4.lkb vector fragment were lic~ated and the recombinant clone (pJ1446) was verified by restriction mapping. This was transformed into the ~ co i KS303 (Strauch, K., et al. (1988) Proc:. Natl. Acad. Sci. USA ~5, 1576) and grown in low-phosphate media (Chang, C. N. (1987) Gene ~5, 189) at 30"C. The receptor fragment protein was purified by hGFI affinity chromatography (Spencer, S. A., et al. (198EI) ,~. Biol. Chem. ~, 7862; Leung, D. W., et al.. (1987) Naturg ~0,, 537) . The sequence for pJ1446 i.s shown in Fig. 12 together with the amino acid sequence: of the cloned receptor.
coli W311~D, degP (Strauch, K. L., et al. (1988) PNAS USfi 85, 1576;1 was transformed with pJ1446 and grown in low-phosphate media (Chang, C. N. (1987) Gene 55, 189) in a fermentor at 30°C. The 246 amino M

acid hGHr wars usef, to generate prel iminary data . A
slightly shorter hGHr containing amino acids 1 through 238 was used in the examples herein. The results obtained with that receptor were indistinguishable from those obtained with the 246 amino acid hGHr.
The plasmid phGHr(1-238) (Table X(A)) was constructed to generate .a stop codon after G1n238 to avoid the problem of carboxyl terminal heterogeneity. The binding protein from KS330 cultures containing phGHr(1-238) was produced in slightly higher yields and with much less heterogeneity (data not shown) than from cultures containing phGHr(1-246).
Routinely, 20 to .40 mg of highly purified binding 1:5 protein coulf, be isolated in 70 to 80 percent yield starting from 0.2 kg of wet cell paste ('2 liters high cell densii~y fermentation broth). Both N-terminal sequencing and peptide mapping coupled to mass spectral analysis of the C-terminal peptide confirmed that the product extended from residues 1 to 238.
Site-directed mut:agenesis of the phGHr (1-246) template was performed (Carter, et al. (1986) Nucleic Acids Res. ~,3, 44:l1-4443) to produce phGHr (1-240, 2'S C241R) using the oligonucleotide 5~-ATG-AGC-CAA-TTT-ACG-CGT-TAG-GAA-GAT-TTC-3':
the asterisks. are mismatches from the phGHr (1-246) template, undlerlined is a new unique MluI site and CGT-TAG directs the C241R mutation followed by a stop codon (Table :K(A)).

Table X(A).
:>equences of amino- and narboxyl-termini of hGH binding protein constructions Plasmid Temini lProtein/DNA sequence/Restriction sites -3 -2 -1 +1 +2 +3 p~hGHr(1-246) Am3.no ALA-TYR-ALA-PHE-SER-GLY
GCC-TAT-GCA-TTT-TCT-GGA
NsiI
phGHr(1-246) Carboxyl :?38 239 240 241 242 243 244 245 246 GLN-PHE-THR-CYS-GLU-GLU-ASP-PHE-TYR-AM
CAA-TTT-ACA-TGT-GAA-GAA-GAT-TTC-TAC-TAG-CGGCCGC
NotI
phGHr (1-240,C241R) Carboxyl Gln-Phe-Thr-Arg-AM
* **
C;AA-TTT-ACG-CGT-TAG-GAA-GAT-TTC-TAC-TAG-CGGCCGC
MluI NotI
phGHr(1-238) Carboxyl C:ln-AM
** **
C;AA-TAG-ACA-CGT-TAG-GAA-GAT-TTC-TAC-TAG-CGGCCGC
NotI
*Indicates mismatches from the gild-type template ~C~~~."~'~4 The plasmid, phGHr (1-238) was produced by site-directed mut~agenesis on the phGHr (1-240, C241R) template using restriction-selection (Wells, et al., (1986) hP il. Trans. R. Soc. Lond. ~, 317, 415-423) !5 against the MluI site (Table X(A)). Briefly, an oligonuc:leotide, 5'-AG-ATG-AGC-CAA-T'AG-ACA-CGT-TAG-GAA-3' introduced a 'translation stop codon after G1n238 (CAA
triplet) and altered the MluI restriction-site (underlined). After growing up the pool of duplex DNA from the initial transfection with heteroduplex, the DNA was restricted with MluI and retransformed to enrich far the desired phGHr (1-238) plasmid prior to DNA sequencing.
It was subsequently determined by DNA sequencing that the cloned hGH binding proteins in phGHr(1-238) contained a '.C51A mutation which arose either as a 2C> cDNA variant or as a cloning artifact. The A51T
revertant: wa:~ therefore to be identical to the published sequence (Leung, et al., (1987) Nature (London) 330, 537-543. The purification and binding properties of the proteins containing either Thr or 2F~ Ala at position 51. were indistinguishable (results not shown). The A1a51 binding protein variant was selected for all subsequent analysis because it had been characterized amore thoroughly.
To compare the specificity of the recombinant hGH
30 binding prote~ln from ~. coli with the natural product isolated from human serum, the affinities were determined for wild-type and various hGH mutants:

~~~~.'~ : 4 Table X(B).
_Ka~'M fS.D. vrotein from' for hGH
binding_ b b m~b H Ka( hGH mutant uman serum Kd wt) E. coli Kd(wt) Kd Encoli) wt 0,.5510.07 - 0.4010.03- 1.4 1:58A 2112 3~Bt6 1411 3615 1.5 F;64A 1211 224 111 285 1.1 F;174A 0.270.1)4 0.4!~0.110.160.01 0.40.1 1.7 F176A 717 13020 485 12020 1.5 '' Values of Kd and corresponding standard deviations (SD) were determined by competitive binding analysis (Fig. 24) with wild-type hGH (wt) and a number of mutants of hGH.
Reduction in binding affinity calculated from the ratio of dissociation constants for the hGH mutant (mut) and wild-type hGH for each hGH binding protein.
Ratio of diss;ociati:on constants for the two hGH binding proteins with a given hGH type.

~~~~.'~'~4 Both proteins formed a specific stoichiometric complex with hGH (Fig. 24). As can be seen, the affinities for wild-type and mutants of hGH are nearly identical between the two binding proteins (right side column, a ra). The recombinant hGH
binding protein has a marginally higher affinity compared. to the natural protein from human serum.
This may reflect t:he greater purity and homogeneity of the recombinant protein. Both proteins had identical sp~ecificities as shown by the changes in binding affinities for four alanine mutants of hGH
that disrupt binding to the hGH binding protein (Kd(mut)/Kd(wt) supra). The affinity of hGH for the binding protean extending to Tyr246 (Kd = 0.36 ~ 0.08 nM) was virtually identical to that terminating after G1n238 (0.40 ~ 0.03 nM) indicating the last 8 residues (including the seventh cysteine in the molecule) are not essential for binding hGH.
Example 4 Receptor and Monoclonal Antibody Bindinct Assay Purified hGH or hGH variants (over 95% pure) were assayed for binding to the soluble hGH receptor of Example 3. Laser dlensitometric scanning of Coomassie stained gels after' SDS-PAGE was used to quantitate 2.5 the concentration of the purified hormones. These values were in close agreement with concentrations determined from the absorbance at 280nm (E2g00.1~ = p,93)~ The dissociation constants (Kd) were calculated from Scatchard analysis for competitive displacement of [125I] hGH binding to the soluble c~GH receptor at 25' C. The 1251 hGH was made according to the method of Spencer, S. A., et al. (1988) T~. Bioc em. 263, 7862.

2~~ 1774 An enzyme-linked immunosorbent assay (ELISA) was used to assess the: binding of eight different monoclonal antibodies to various segment-substituted and residue-substituted hGH variants. The following are the Mabs used:
'jab :Cdenti~ty _, Source/Method 1 MabA (*) 2 33.:> Hybritech, Inc.

3 Cat~~ H-29!3-O1 Medix Biotech, Inc.

4 72.:: Hybritech, Inc.

5 Cat~ H-299-02 Medix Biotech, Inc.

6 Mab 653 Chemicon 7 Mab D (*) 8 Mab B ( * ) (*) Carbone, F. R.,. et al. (1985) J. Immunol. 135, Rabbit polycl~~nal antibodies to hGH were affinity purified and coated onto microtiter plates (Nunc plates, InterMed; :Denmark) at 2 ~g/mL (final) in 0.005 M sodium carbonate pH 10) at 24°C for 16-20 h.
Plates were :reacted with 0.1 ~g/mL of each hGH
variant i.n buffer 13 (50 mM Tris [pH 7.5], 0.15 M
NaCl , 2 mM ED'.CA, 5 mg/mL BSA, 0 . 05 % Tween 2 0 ; 0 . 02 %
sodium azide) for i:wo hours at 25°C. Plates were washed and tr~en incubated with the indicated Mab Which was ser~:ally diluted from 150 to 0.002 nM in buffer H. ~~fter two hours plates were washed, stained with horseradish peroxidase conjugated anti-mouse antibody and assayed. Values obtained represent: th~= concentrations (nM) of each Mab necessary to produce half-maximal binding to the respective hGH variant.
* Trade-Mark .,.

~C~~~.'~'~4 Competitive displacement of the hGH receptor from hGH
by anti-hGH rsabs was determined as follows. Assays were carried out b~y immobilization of wild-type hGH
in microtite:r plates coated with anit-hGH rabbit !5 polyclonal antibodies as described. Receptor (fixed at 10 nMj and given anti-hGH Mab (diluted over a range of 150 to 0.002 nMj were added to the hGH
coated microtiter plate for 16-20 hours at 25°C, and unbound components were washed away. The amount of bound receptor was quantified by adding an anti-receptor Mab that was conjugated to horseradish peroxidase wlhich did not interfere with binding between hGH and the receptor. The normalized displacement value was calculated from ratio of the concentration of Mab necessary to displace 50% of the receptor to the half-maximal concentration of Mab necessary to saturate hGH on the plate. This value was used to compares the relative ability of each Mab to displace the receptor.
2C1 Example 5 Active Domain: for ~Somatoctenic Receptor Binding The 17 segment sub~~tituted hGH variants described in Example 1 and Example 2 were assayed for binding to the soluble ~somatogenic receptor of Example 3 and 2~~ binding to the monoclonal antibodies as described in Example 4. z'he results of the binding assay to the somatogenic receptor is shown in Table III. As can be seen, the segment substitutions that are most disruptive to binding are within regions A, C and F
30 of Figs. 4 and 5. These regions were further directed into smaller segments to further localize the active domains of the hGH molecule involved in binding to t:he somatogenic receptor. The most significant results from Table III are shown in Fig.
35 4 which is a t>ar graph showing the relative reduction ~~'~~~'~'~4 in binding to 'the soluble hGH receptor as a consequence of the substitution of the indicated analogous sequences from the analogs hPRL, hPL and pGH as shown. Three active domains were identified as regians A, C and F comprising amino acid residues 12-19, 54-74 and lE>4-190 respectively. These regions are identified in the three-dimensional representation of t:he hGH molecule in Fig. 5.
As can be seen, thsa three active domains, A, C and F, although discontinuous in the amino acid sequence of hGH, form a continuous region in the folded molecule which defines the ::omatogenic binding site on hGH.
Example 6 E~itope Ma_ppincr of hGH
The binding of the eight different monoclonal antibodies to specific segment-substituted hGH
variants is shown i.n Table XI.

;~~~ ~~,'~'"i 4 TABLE XI
Mab 1 2 _ 3 4 5 6 7 8 Hybr Medix HybrMedix InGH Variant MCA 33.2 1 72.32 Chemicon MCD MCB

pat hGH 0.4 0.4 0.1 0.050.2 0.2 0.08 0.1 hPL(12-25) 0.4 0.4 >75 >50 0.2 0.2 0.08 0.1 hGH(11-33) 0.4 >100 1.5 0.050.2 0.2 0.08 0.1 llPRL(12-33) 0.4 >100 >75 >50 0.2 0.2 0.08 0.1 hPRL(12-19) 0.4 >12 >75 >50 0.2 0.2 0.08 0.1 hPRL(22-33) 0.4 0.4 0.1 0.050.2 0.2 0.08 0.1 hPL(46-52) 0.4 0.4 0.1 0.050.2 0.2 0.40 0.1 1~GH(48-52) 0.4 0.4 0.1 0.050.2 0.2 0.08 0.1 hPL(56-64) 0.4 0.4 U.1 0.050.2 0.8 0.08 0.1 lGH(57-73) 0.4 0.4 0.1 0.05>200 >200 0.08 0.1 hPRL(54-74) 0.4 0.4 0.1 0.050.2 0.6 0.08 0.1 hPRL(88-95) >400 0.4 0.1 0.050.2 0.2 0.08 0.1 hPRL(97-104) >400 >12 0.1 0.050.2 0.2 0.08 0.1 hPL(109-112) >12 0.4 >75 15 0.2 0.2 0.08 0.1 hPRL(111-129) >12 0.4 >75 >50 0.2 0.2 0.08 0.1 hPRL(126-136) 0.4 0,4 0.1 0.050.2 0.2 0.08 0.1 pGH(164-190) 0.4 0,4 0.5 0.3 >25 12.5 0.20 0.4 pGH(167-182) hGH(~32-46) 0.4 0.4 0.1 0.05 0.2 0.2 >100 >100 P~12A 0.4 0.4 >75 >50 0.2 0.2 0.08 0.1 C182A 0.4 0.4 0.1 0.05 2.0 0.2 0.08 0.1 With the possible exception of the pGH (167-190) variant, disruption of binding to each monoclonal antibody was dramai~ic and highly selective. Figures 13 through 20 localize the epitope for each of the Mabs on the three-dimensional structure of hGH. Fig.
6 comprises 'these epitopes to the binding site for the somatogenic receptor.
For example, the hPRL (88-95), hPRL (97-104), hPL (109-112) and h,PRL (111-129) variants do not bind to Mabl yet: the other segment-substituted hGHs outside of these regions bound as effectively as wild-type hGH. Binding to Mabs 2, 3, 4, 5 and 6 was disrupted by mutations in discontinuous regions in the primary sequence but in close proximity in the ~~~~.~~I

fclded hormo~ae (se:e Figs. 6 and 14 through 19) . In contrast:, Mabs 1, '7 and 8 were disrupted by mutations defined by a, continuous sequence as shown in Figs.
13, 19 and 20.
The regions disrupting binding to a given monoclonal antibody werE~ further analyzed by directing specific segment-substituted hGH variants into subdomains or by analyzing variants that had common substitutions that stall bound to the particular Mab. For example, pGH (11-33) retained tight binding to Mab 4 yet hPRL (1.2-33) disrupted binding. Thus, the disruptive mutations in the hPRL (12-33) variant can be confined 'to residues not mutated in pGH (11-33):
N12, L15, R1E>, D26 and E30. This set can be further restricted to N12, L15 and R16 because the hPRL
(12-19) varieint disrupts binding, but the hPRL (22-23) variant does not (see Fig. 16). The N12H
mutation in hPL (12-25) can entirely account for the disruption in binding to Mab 4 because this is the only mutation not in common with pGH (11-33). This was tested by substituting alanine for Asn-12. The binding of Mabs 3 or 4 to the N12A residue-substituted hGH variant was reduced by over 100-fold whereas :binding to the other Mabs was uneffected.
2.5 Using this set of hGH variants, it was possible to resolve the e:pitopeas from all eight Mabs even though binding for most of these Mabs was blocked by a common set of mutations. For example, although hPRL
(12-19) disrupted lbinding to Mabs 2, 3 and 4, other variants indicatead that these Mabs recognized different st~:-uctur~es. Specifically, Mabs 2 and 3 were blocked by pGH; (11-33) yet Mab 4 was uneffected.
Binding of Mabs 3 and 4 was blocked by hPL (12-25) yet binding to Mab 2 was uneffected. Thus, the eight ~~~~.~!

antibodies may have epitopes that overlap but none superimposed. Muitations that disrupt binding are present in both helices and loops and are always in close proximity in the folded hormone.
:5 Collectively, the ~epitopes with a set of eight Mabs cover most of the ihormone. However, there are still regions where these Mabs did not bind. For example, three of the 20 variants did not significantly disrupt binding to any of the Mabs tested (hPRL (22-33), pGH (48-52) and hPRL (126-136)).
There are significant differences between the antibody epitopes and the receptor binding site.
Firstly, the patch defined by disruptive mutations is larger for the receptor than for any of the Mabs.
1!5 A seconf, difference is that the receptor has more tolerence to disruptive substitutions in the hormone than do the Mabs. This is evidenced by the fact that the maximum reduction in binding to the receptor for any of the mutants is about 70-fold, whereas almost every antibody has at least one variant that causes more than a 1000-fald reduction in binding some of which may be the result of single substitutions such as N12A.
example 7 2 'i Competitive B.i.ndinct of Mabs and shGHr Many of the variants which cause disruption of receptor binding al'.so disrupt the binding of one or more of the riabs. The ability of each of the eight Mabs to block the binding of the hGH receptor to hGH
was therefore evaluated. Results of this assay are shown in Table XII.

~~~~~.'~'~4 TABLE XII
Mab 50X bindingdisplace 50X Normalized displacement to hGHt of receptor ~conc. for 50X displacement conc. for 50X binding 1 0.4 >150 >375 2 0.4 0.8 2 3 0.1 150 1500 4 0.05 150 3000 0.2 0.2 1 6 0.2 0.2 1 7 0.08 0.4 5 8 0.1 >150* >1500 ~*~ Binding of Mab 8 appears to slightly enhance binding of receptor to hGH.
tData from Table X for 'binding of each Mab to hGH.
As can be seen Mab~: 5 and 6 are the most efficient at blocking binding of the hGH receptor. This is because these M;abs have antigenic determinants located in the loop from residues 54 through 74 and 5 in helix 4 closely overlap determinants for the receptor (see Figs,. 5, 6, 17 and 18). Mab 2 was the next most competitive antibody and it too shared a common disruptive mutation with the receptor (hPRL
(12-19)). In cantrast, Mabs 3 and 4 were roughly 1000-fold less competitive than Mab 2 yet they also shared overlapping disruptive mutations with the receptor in helix 1. See Figs. 15 and 16. This apparent discrepancy may be easily reconciled if the mutations in helix I that disrupt Mabs 3 and 4 differ from those residues. which disrupt binding to Mab 2 or the receptor. :Indeed, one such mutant (N12A) disrupts binding of either Mab 3 or 4 without effecting binding t.o Mab 2 or the receptor.

~r~~~.~~f Mab 7 competes relatively strongly with the receptor for hGH and it is. disrupted by segment-substituted hGH variant:: that cause a minor disruption of receptor bir,~ding, e.g., hPL (46-52). Thus, it appears that Mabs 2 and 7 sit on the border of the receptor binding site. Mabs 1 and 8 were unable to give det.ectat>le di;splacement of the receptor, and as expected. ths~se contain no overlapping antigenic determinants with the receptor. These competitive binding data taken together with the direct epitope mapping and receptor binding data strongly support the general location of the receptor binding site as shown in Fig. 5.
Example 8 Rector Active Amino Acid Residues The analysis of hGH in Examples 5, 6 and 7 implicate the amino terminal portion of helix 1 (residues 11-19) as being oi° moderate importance to receptor binding. In addition, residues 54-74 and 167-191 were identified as being important to receptor binding. Identification of which amino acids in these damains: which are active in receptor binding was carried out b!~ analyzing a total of 63 single alanine variants. See Tables XIII, XIV and XV.

'~~~~.'~'~4 _77_ TABLE XIII
Amino acid scanning of positions 2-19 in hGH
Variant Kd(nM) Kd(variant)/Kd(wt) wt 0.34 1.0 P2A. 0.31 0.90 T3A. 0.31 0.90 I4A 0.68 2.0 P5A 0.71 2.1 L6A 0.95 2.8 S7A 0.61 1.8 R8A 0.48 1.4 L9A 0.32 0.95 F10A 2.0 5.9 N12A 0.40 1.2 A13(WT) M14A 0.75 2.2 L15A 0.44 1.3 R16A 0.51 1.6 A17 (WT) H18A 0.24 0.71 R19A 0.37 1.1 ~~t~~.'~'~4 _78_ TABLE XIV
Amino an~3 scanning of positions 54-74 in hGH
Variant Kd(nM) Kd variant/Kd WT

WT 0.31 1.0 F54A 1.5 4.4 S55A 0.41 1.2 E56A 1.4 4.1 S57A 0.48 1.4 I58A 5.6 17.0 P59A 0.65 1.9 S62A 0.95 2.8 N63A 1.12 3.3 R64A 7.11 21.0 E65A 0.20 0.6 E66A 0.71 2.1 T6'7A NE -Q6~BA 1.8 5.2 Q6'9A 0.31 0.9 K70A 0.82 2.4 S7:lA 0. 68 2 . 0 L7:3A 0.24 0.70 ~f~~~.'~"~4 TABLE XV
Amino acid scanning of positions 167-191 in hGH
Variant Kd(nM) Kd variant/Kd WT

WT 0.34 1 R167A 0.26 0.75 K168A 0.37 1.1 M1'7 OA NE -D1'71A 2 . 4 7 . 1 Kl'72A 4 . 6 14 V1'73A NE -E1'74A 0.075 0.22 T1'75A NE -T1'75S 5. 9 16 F1'76A 5 . 4 16 L1'7 7A NE -+

R1'78A NE -R1'78N 1.4 4 . 2 I1'79A 0.92 2.7 V180A 0.34 1.0 Q181A 0.54 1.6 C182A 1.9 5.7 R183A 0.71 2.1 S184A 0.31 0.90 V185A 1.5 4.5 E186A 0.27 0.80 G187A 0.61 1.8 S188A 0.24 0.7 _ F191A 0.20 0.60 The substitution of alanine was extended to include residues 2-:l9 because of uncertainties in the position of the amino terminal residue (Abdel-Meguid, S . S . , et al . ( 198 7 ) Froc . Natl . Acad . Sci . USA ~4 , 6434). Indeed, the mast pronounced reduction in binding occurred for F10A (6-fold) followed by alanine substitutions at residues 4-6 at the N-terminus of lhelix 1 (see Fig. 21). Substantially larger effecta on binding (greater than 20-fold) E~~.''7''74 occurred for specific alanine substitutions within the 54 to 74 loop and the carboxy terminal sequence 167-191. For several alanine variants, binding was enhanced up to 4. °_.-fold. The most dramatic example was E174A which was located in the midst of a number of disruptivE~ alanine mutations. Sees Fig. 4, 7 and 21.
The most disruptive alanine substitutions form a patch of about 25A by 25A on the hormone that extends from F1G to 1264 and from D171 to V185 (see Fig. 21) .
Furthermore, these side chains appear to be facing in the same direction on the molecule. For example, all of the alanine muitants that most effect binding on helix 4 (D171A, K172A, E174A, F176A, I179A, C182A and R183A) are confined to three and one-half turns of this helix, and their side chains project from the same face of the helix (see Fig. 21). Based upon this model, it Wi3S predicted that T175 and 8178 should be involved in binding because they occupy a central position a~; shown in Fig. 21.
Although the T175A mutant could not be expressed in high enough ~iields in shake flasks to be assayed, a more conservative :mutant (T175S) was. Accordingly, the T175S mutant caused a 16-fold reduction in 2:5 receptor binding. Similarly, although R178A was poorly expressed, R178N could be expressed in yields that permitted analysis. R178N exhibited a greater than four-fold reduction in binding affinity.
The next most disruptive mutant in the carboxy 3i~ terminus was V185A.. Although V185A is outside of helix 4, it is predicted by the model to face in the same direction as the disruptive mutations within helix 4. In contrast, alanine mutations outside the ~~,~~~.'~'~4 binding patch, or within it facing in the opposite direction from those above (R167A, K168A, V180A, Q181A, S184p,, E186A, S188A) generally had no or little effect: on reaceptor binding.
A similar analysis applied to alanine mutants in helix 7., albeit with more moderate effects on binding. Within the helix, the alanine substitutions that most disrupted binding were at residue 6, 10 and 14 which werEa located on the same face of the helix.
The least di~;ruptiVe alanine mutations (L9A, N12A and L15A) were l~~cated on the opposite face of helix 1.
This is further confirmed by the fact that anti-hGH
Mabs 3 and 4 which do not compete with the receptor for binding vto hGH, both bind to Asn-12. See Table XVI.

~:~w~~.'~'~4 TABLE XVI
Binding of hGH and alanine variants to eight different ant i-hGH monoclonal antibodies (Mab).
Mab Hormone1 2 3 4 5 6 7 8 hGH 0.4 0.4 0.1 0.05 0.2 0.2 0.08 0.1 F10A 0.4 0.4 0.1 0.05 0.2 0.2 0.08 0.1 N12A 0.4 0.4 >75 >50 0.2 0.2 0.08 0.1 I58A 0.4 0.4 0.1 0.05 0.2 0.2 0.08 0.1 R64A 0.4 0.4 0.1 0.05 0.2 1.6 0.08 0.1 Q68A 0.4 0.4 0.1 0.05 0.2 0.2 0.08 0.1 K168A 0.4 0.4 0.1 0.05 0.2 0.2 0.08 0.1 D171A 0.4 0.4 0.1 0.05 0.2 0.2 0.08 0.1 K172A 0.4 0.4 0.1 0.05 0.2 0.2 0.08 0.1 E174A 0.4 0.4 0.1 0.05 0.2 0.2 0.08 0.1 F176A 0.4 0.4 U.1 0.05 0.2 0.2 0.08 0.1 C182A 0.4 0.4 0.1 0.05 2.0 0.2 0.08 0.1 V185A 0.4 0.4 0.1 0.05 0.2 0.2 0.08 0.1 The relative positions of side chains within the 54-74 loop cannot be fixed in the model as they can be for those within helices 1 and 4. However, there is a striking pe:riodic:ity in the binding data in which mutations of even numbered residues cause large reductions i:n binding relative to odd numbered residues" This is especially true for the first part of this region (54-59) and may reflect a structure in which even reaidue:~ project toward the receptor and 1() odd ones away.

~:~~~.'~"~4 Example 9 Conformational Ints~grity and Binding Energetics of Alanine Substituted hGH Variants Several liner of evidence indicate that the alanine substitutions that disrupt the receptor binding do not do so by causing the molecule to be misfolded.
Firstly, the eight Mabs react as well with almost all of the alanine mutants that disrupt binding to the receptor as they dc. with hGH. See Table XII supra.
The exceptions are R64A and C182A which selectively disrupt binding t.o the anti-hGH Mabs 6 and 5, respectively. These two Mabs as previously indicated compete with the somatogenic receptor for binding to hGH. In addition, two alanine variants were made 1!5 which do not effect receptor binding. One of these effects the r~indinc~ of two Mabs (N12A) and the other effects none of the Mabs (K168A). This data indicates that Minding to either the Mabs or receptors is disrupted by a very local pertabation in the structure of the variant. Moreover, the far W
circular dichroic spectra of all the hGH variants tested are virtually identical to wild-type hGH.
About 20% of the alanine mutants (D11A, T60A, P61A, T67A, N72A, E74A, I)169A, M170A, V173A, T175A, L177A, K178A, C189A, G190A) were not secreted at high enough levels in sh<ike flask to be isolated and analyzed.
Since genes encoding such variants were expressed in the same vect~~r and expression was independent of the specific alanine c.adon, variations in steady-state expression levels most likely reflect differences in secretion level and/or proteolytic degradation of the hGH variants. Sevearal of the non-expressing alanine variants in helix 4 are located on its hydrophobic face (M170A, V173A and L177A) as shown in Fig. 21 ~~~~.'~"~4 wherein the hydrophobic side of the helix is identified by~ open shading. However, this is not a general effects because several alanine substitutions were toleratEad on the hydrophobic face of helix 1 (L6A, L9A and F10A) and helix 4 (F176A and V180A).
In addition, impair.~ed expression of hGH variants was sometimes observed when charged or neutral amino acids were rceplace~d with alanine (D11A, T60A, T67A, N72A, E74A, Lr169A, T175A, R178A). Mutations such as In T175S and R1T8N, that preserved the hydrogen bonding group at. those sites, could be expressed albeit at levels below wild--type. The non-expressing C189A
variant disrupts the carboxy-terminal disulfide and its counterpart (C182A) was also expressed at levels la far below wild-type. Several other non-expressing alanine mutants (T60A, T61A and T67A) were located in a loop structure. Thus, low levels of expression or non-express ion ca.n result from a multitude of structural effects but can be obviated by isosteric 20 or isofunctio:nal substitutions.
The substitutions 'that cause a ten fold or greater effect upon binding (I58A, R64A, K172A, T175S, F176A) are likely tc~ be directly involved in binding. The strengths of hydrogen bonds or salt bridges present 25 in nature (Fersht, A. R. (1972) J. Mol. Biol. ~4_, 497; Brown, L. R., et al. (1978) Eur. J. Biochem. ,~8, 87; Malivor, R. , et: al. (1973) J. Mol. Biol. ~ 123) or engineered k>y site-directed mutagenesis experiments (iFersht, A. R. , et al. (1985) Nature ~, 30 225: Bryan, F'. , et al. (1986) hoc. Natl. Acad. Sci Z~SA ~, 3743; Wells, J. A., et al. (1987) proc. Natl.
Acad. Sci USA; ~4,, 5167; Wells, J. A., et al. (1987) lProc. Natl. Acad~Sci. USA 84, 1219; Cronin, C. N.,et al. (1987) J. Am. Chem. Soc. X09 2222: Graf, L., et ~~~~.""~ r 4 al. (1988) Proc. Natl. Acad. Sci. USA 85 4961) overlap and range widely from 1 to 5 kcal/mole depending upon the microenvironment. For hGH, reductions in binding fee energy of 0.8, 1.0, 1.2, 1. 6 and 1. a3 kcal/mol (eeGbinding - +RT 1n Kd (var)/Kd(wt)) occurred for alanine substitutions at E56, Q68, D~171, K172 and R64, respectively. The energetics for burial of a hydrophobic side chain into a protein tends to parallel its free energy of transfer into ethanol (Estell, D. A., et al. (1986) Science 233, 659; Nozaki, Y. et al. (1980) in The Hydrophobic Effect (wiley, N.Y.. pp. 4-21).
Accordingly, the reductions in binding free energies for F175A, F10A, F°_i4A, I58A, and V185A were 1.6, 1.0, 0.9, 1.7 and 0.9 l~;cal/mol, respectively. These are slightly below the' predicted change in hydrophobic free energy :in going from Phe, Ile or Val to Ala of 2.0, 2.4 and 1.0 kcal/mol, respectively. By this analysis the effect: of the T175S mutant (eeGbinding -l.6kca1/mol) is larger than expected for loss of a y methyl group (eeGrlydrophobic - 0.7 kcal/mol). To fully characaerize~ the nature of the molecular contacts between hGH and its somatogenic receptor requires direct structural information. However, the 2 5 energetics of binding of these alanine mutants shows them to be in the range of previous measurements made on contact re=sidues in entirely different systems.
In fact, the sum of binding free energies for these alanine-substituted variants exclusive of C182A that 3~D are most disruptive to receptor binding (-13.2 kcal/mol.) ifs comparable to the total free energy binding between hGH: and its receptor (-13 kcal\mol).

'~~~1."~ r ~~

Example 10 Reactivity of hGH ~~ariants with Anti-hGH Polv~clonal. Antibodies The hGH variants hPRL (22-33), E174A and hPRL (88-95) were tested :gin a rat weight gain assay. The results of that assay are presented in Fig. 22. As can be seen, all the variants except hPRL (22-33) have a reduced potency after about 14 days of growth. The leveling off of growth is attributed to the development of antibodies to the various growth hormones which neutralize the biological effect. The fact that the hPRL (22-33) variant continues to induce growth, suggests that it is not as immunogenic as wild-type hGH or the other variants used.
A comparison. of the reactivity of various hGH
variants with human and murine serum containing polyclonal antibodies to hGH is shown in Table XVII.

~~~~.'~"~4 TABLE XVII
Serum Anti-IzGH Antibodies Binding to hGH Variants Average X. of Reduction of Anti-Protropin Binding ~ SD X Incidence Human Mouse Human Mouse Sera Sera a a Sera N-22, (N-6) hGH 0 0 100 100 pGH 11-33 86 t 13 65 t16 100 100 hPRL 12-33 79 t 19 52 t13 100 100 hPL 12-25 35 t 19 16 11 81 33 hPRL 12-19 29 20 11 t12 71 33 hPRL 22-33 69 15 38 8 100 100 hPL 46-52 6 8 2 4 10 0 pGH 48-52 7 8 4 4 10 0 pGH 57-73 43 15 39 f12 95 100 hPRL 54-74 14 9 8 7 24 0 D80 13 15 7 f7 14 0 hPRL 88-95 14 22 4 5 19 0 hPL 109-11210 t 11 9 9 24 17 hPRL 126-1368 12 2 2 19 0 C182A 1 5 1 t3 5 0 As can be ss:en, variants containing substitutions within the region from residues 22 to 33 have substantially reduced binding activity, and in some cases no activity, with individual human and mouse anti-serum fox- wild-type hGH.
Except far thee variant pGH 57-73, variants containing substitutions in the other regions shown do not have a significant reduction in reactivity. Since the segment substituted mutants between residues 11 and 33 retain their ability to bind the somatogenic receptor, such variants demonstrate the production of variants which maintain the ability to promote somatogenesis but have another property which is -88_ modified, in this. case reactivity with anti-hGH
polyclonal antibodies.
Example 11 Relationship Between I~C and Potency 'S A semi-log plot of the ratio of Kd (variant)/Kd (wild type) fo:r specific hGH variants versus the potency of such variants in a rat weight gain assay is shown in Fig. 23. As can be. seen a linear relationship exists which suggests that a decreased-binding affinity for the somat.ogenic: receptor will result in decrease in poten~.y.
As can be seen, the hGH variant E174A has a higher binding affinity for the somatogenic receptor than the wild-type hGH. Its potency is also greater than 1!5 that of wild-type hGH by about 12%.
Further, the variant pPRL (97-104) has essentially the same binding constant as wild-type hGH but about a 2.7-fold in~~rease in potency.
Example 12 Active Domains in hGH for Prolactin Receptor Binding Human growth hormone (hGH) elicits a myriad of physiological effects including linear growth, lactation, nitrogen retention, diabetogenic and insulin-like ~=ffects, and macrophage activation. R.K.
Chawla, J.S. lParks .and D. Rudman, Annu. Rev. Med. 34, 519-547 (1983): O.G.P. Isaksson, et al. (1985) nu.
~tev. Physiol. ~7, 483-499; C.K. Edwards, et al., (1988) cience_ ~9_, 769-771. Each of these effects begins with the interaction of hGH with specific cellular receptors. J.P. Hughs, et al. (1985) nu.
~tev. Physiol" 47, 469-482. Thus far, the only cloned genes whose products bind hGH are the hGH

~s receptor from liver (D. W. Leung, et al., (1987) Nature (London) 330, 537-543) and the human prolactin (hPRL) receptor from mammary gland (J.M .Boutin, et al., (1988) Cell 53, 69-77).
Receptor "sp:illover" of hGH onto the hPRL receptor has clinical precidence in cases where acromegalics, who producca high levels of hGH, develop a hyperprolactinemic syndrome despite having normal levels of hF~RL (J' . E . Fradkin, et al . , ( 1989 ) T~ew Engl. J. Med. 320, 640-644). However, other receptors exist i:hat bind hGH , including the placental lactogen (PL) receptor (M. Freemark, et al . , ( 1987 ) Endocrinolocrv 120, 1865-1872 ) . It previously was not known if the binding sites on hGH
for these re~~eptors are identical or which receptor (or combination of receptors) is responsible for which pharmacological effect. To begin to address these issues the hGH and hPRL receptor binding sites on hGH were mapped. The results obtained indicate that these receptor binding sites overlap but are not identical. 'his has allowed the rational design of receptor specific variants of hGH.
The hGH and hPRL receptors contain an extracellular hormone binding domain that share 32% sequence 2!5 identity, a sing7Le transmembrane domain, and a cytoplasmic dlomain which differs widely in sequence and length. 'The extracellular binding domain of the hGH receptor has been expressed in E. coli and has identical binding F>roperties to that found naturally as a soluble serum binding protein (S.A. Spencer, et al., (1988) 7~ Biol. Chem. 63, 7862-7867).
Similarly, t:he extracellular domain of the hPRL
receptor has been expressed in E. coli and purified.
The hPRL receptor fragment extends from residues Glnl 3!5 to Thr211 and terminates just before the single Aur ~ ~~ ~~~

transmembrane domain. It retains high binding affinity and specii:icity that is virtually identical to its full-length receptor. The gene encoding the hPRL receptoo used in the experiments was kindly provided by I)r. P.~A. Kelly, Laboratory of Molecular Endocrinology, McG:ill University, Montreal, Canada.
This DNA sec~aence was obtained from a human mammary cDNA library and identified with a probe covering known conserved regions amongst cross-species members l0 of the prolactin rEaceptor family. See _e. g., Davies, J.A., et al., (1989) Mol. Endrocrinologv 3, 674-680:
Edery, et al. (1989) Proc. Natl. Acad. Sci. USA 86 2112-2116; Jolicoeur, et al. (1989) Mol.
Endrocinology _3, 895-900. These truncated and highly 1_°°> purified receptors are extremely useful reagents for rapid and accurate assessment of binding affinity for mutants of hGl~i.
Relationship between hPRL and hGH receptor binding sites.
20 To determine if the epitopes for the hGH and hPRL
receptors overlapped we analyzed whether or not the hPRL receptor fragment could displace the hGH
receptor fragment from hGH (results not shown).
Indeed, the hPRL receptor fragment competed for the 2F~ hGH receptor binding site with an apparent Kd of 1 nM. This is virtually the same affinity as that measured by direct binding of the hPRL receptor to hGH (results not shown).
Eleven o.f the: segment-substituted hGH variants from 30 Table III were used to localize the epitope on hGH
for the hPRL receptor. The hGHn32-46 variant was also used in this experiment. The approach was similar to that used to determine the epitope on hGH
for the hGH receptor as previously described i.e. by ~~Q~.'~'~4 the disruption in binding of variants of hGH except that the receptor was hPRLr rather than hGHs. The results for t:he above twelve segment-substituted hGH
variants are summarized in Table XVIII.
Table XVIII.
Binding of hGH variants produced by homolog-scanning mutagenesis t:o the extracellular domain of the hPRL
receptor (hPRLr) . Mutants are named according to the extremes of segment substituted from the various hGH homolog;s: pGH, hPL, or hPRL. The exact description of the mutations introduced is given by the series of single mutants separated by commas.
The component: single mutants are designated by the single letter code for the wild-type residue followed by its codon position in mature hGH and then the mutant residue. Mutants of hGH were produced and purified as previously described herein. Binding to hPRLr was measured essentially as described for the hGHr (Spenceo, S.~~. et.al. (1988) J. Biol. Chem.
~63,7862~-7867) except that affinity purified rabbit polyclonal antibodies raised against the hPRLr were used to precipitates the hPRLr complex with Gibco BSA
(crude) as carrier protein. Standard deviations in values of Kp were typically at or below 20% of the reported value. The relative reduction in binding affinity (Kp~;mut)/1Kp(hGH)) for the hGHr was taken from Table III herein. The change in receptor preference was calculated from the ratios of the relative reductions, in binding affinity for the hGHr to the hPRLr. WT = wild-type.
Change in receptor hPRLr hGHr preference Nament Introduced KD(~) KD(hGH) Kp(hGH) hPRLr WT hGH none 2.3 (1) (1) (1) pGH

(11-33)D11A, M14V, 852 370 3.4 110 H18Q, R19H, F25A, Q29K, E33R

pGH

(48-52)P48A, T50A, S51A, 2.0 0.9 2.8 0.32 Table ontinued) XVIII.
(C

pGH

(57-73) S57T,T60A, 167 73 17 4.3 S62T, N63G,R64K, E65D, T67A,K70R, N72D, hGH

(~32-46) Deler_ion of 14 6.1 ND

residues to hPL

(46-52) Q46H,N47D, 4.4 1.9 7.2 0.26 P48S, Q49E,L52F

hPL

(56-64) E56D,R64M 4.1 1.8 30 0.06 hPRL

(12-19) N12R,M14V, 3.2 1.4 17 0.08 L15V, R16L,R19Y

hPRL

(22-33) Q22N,F25S, 168 73 0.85 85 D26E, Q29S,E30Q, hPRL

(54-74) F54H,S55T, 2.5 1.1 69 0.02 E56S, I58L,P59A, S62E, N63D,R64K, E66Q, T67A,K70M, S71N, N72Q,L73K, hPRL

(88-95) E88G,Q91Y, 3.8 1.6 1.4 1.1 F92H, R94T,S95E

hPRL

(97-104) F97R,A98G, 12.1 5.2 1.6 3.2 N99M, 100c~, , V102A,Y103:P, hPRL

(111-129) Y111'J,L113:I, 2.6 1.1 1.5 0.73 , , E1181C,E119R, G120:L,Q122;E, T123G,G126:L, 8127 E129;5 I
, t.TT L. DDT .",ro ~ ~ ~ ~i -~ nn nnn ~::~~.~~I

As can be seen pGH (11-33) and pGH (57-73) cause large disruptions :in hPRL receptar binding affinity, whereas pGH (48-52) has no effect. Unlike the hGH
receptor, the hPRL receptor will bind hPRL and hPL
but not pGH" As expected, virtually all of the substitutions tested from the binding competent hormones, hPF~L or :hPL, did not disrupt binding. The only exception saws hPRL (22-33) which caused a >70-fold reduction in binding affinity for the hPRL
receptor. Thus, t;he hPRL receptar is very sensitive to mutations in hGH near the central portion of helix 1 and the loop between residues 57 and 73.
The homolog-scan data also suggest that the hPRL and hGH receptor epit:opes are not identical because several segrnent substituted variants cause huge changes in receptor binding preference (Table XVIII).
For example, the disruption in binding caused by the pGH (11-33) or hPRL (22-33) are about 100-fold greater for the hPRL receptor than for the hGH
receptor. J~:n contrast, the hPL (56-64) and hPRL
(54-74) have almost no affect on the hPRL receptor, whereas they weaken binding to the hGH.receptor by factors of 17 .and 69, respectively. These preferential binding Effects (along with binding of monoclonal antibodies as previously discussed) further substantiate that reductions in receptor binding affinity are caused by local and not global structural changes in the mutants of hGH.
The specific side-chains in hGH that strongly modulate binding to the hFRL receptor were identified by alanine-scanning mutagenesis and homologous substitutions.. The hGH varients shown in Table XIX
were prepared.. Then hPRL substitutions, F25S and D26E
cause the largest reductions in binding affinity (21 f~~~,"~'~4 and 4.5-fold,, respectively) in helix 1. These residues project from the hydrophilic face of helix 1 (Fig. 25B) and a:re on the same side as other mutations in helix 1 (notably H18A and F10A) that have milder e:Efects on binding.
Four residues in the loop region (54 to 68) known to affect binding of hGH receptor as well as two residues (Q49A and T50A) preceding this region that are nearby and do not affect hGH receptor binding were tested. The most disruptive mutants are I58A
and R64A which reduced binding affinity by 32 and 6-fold, respectively; the other four mutations have negligible ef:Eects.
The fact thai:. helix 1 and the loop region (58-64) 1_°°> contain strong binding determinants for the hPRL
receptor, imF~licate: helix 4 because this helix is wedged between these two structures (Fig. 25B).
Indeed, alanine-scanning of the helix 4 region between a disulfide linked to C165 through V185 reveals strong binding determinants (Table XIX). The most disruptive mutations extend nearly four helical turns, from 8:167 to 8178, and are located on the same hydrophilic face.
Table XIX.
Binding of single mutants of hGH to hPRL or hGH
receptor fragments (hPRLr or hGHr). Mutants of hGH
were prepared. and purified as previously described except for Q22N, F25S, D26E, Q29S and E33K which were produced by site-directed mutagenesis (Cunningham, B.C. and Wells, J.,A. (1989) Science ~4, 1330-1335:
Zoller, M.J. and Smith, M. (1982) Nucleic Acids Res.
~, 6487-6499). Recector binding assays and mutant nomenclature are described in Table XVIII. Data for the reduction in :binding affinity to the hGHr is taken from Table III. ND indicates not determined.

~~~1~~4 Tab7.e XIX. (Continued) Change in receptor - hPRLr hGHr preference Mutant KD(~) KD(hGH) KD(hGH) hPRLr WT hGH 2.3 (1) (1) (1) P2A 1.3 0.6 0.9 0.7 T3A 3.4 1.5 0.9 1.7 P5A 2.5 1.1 2.1 0.5 L6A ~4.0 1.8 2.8 0.6 S7A 1.9 0.8 1.8 0.4 F10A .g.l 3.5 5.9 0.6 N12A 1.9 0.8 1.2 0.7 M14A 1.3 0.6 2.2 0.3 L15A 1.2 0.5 1.3 0.4 H18A 3.9 1.7 1.6 0.6 R19A 1.4 0.6 0.7 2.4 Q22N 2.1 0.9 ND -F25S 4;B 21 ND -D26E 1~0 4.5 ND -Q29S 3.2 1.4 ND -E33K 1.8 0.8 ND

Q49A 1.5 0.7 ND

T50A 1.9 0.8 ND -F54A 1.8 0.8 4.4 0.2 I58A 73 32 17 1.9 R64A 13 5.7 21 0.3 Q68A 3.1 1.2 5.2 0.3 R167A '7.4 3.2 0.75 4.3 K168A 5.B 25 1.1 23 D171A 3.6 1.6 7.1 0.2 K172A 143 62 14 4.4 E174A 5'9 26 0.22 120 F176A 12'9 56 16 3.5 R178N 2.4 1.0 8.5 0.1 R178K ~6.7 2.9 ND -I179M 1.3 0.6 2.7 0.2 V185A 3.9 1.7 4.5 0.4 Functional contour maps were derived based upon the location of t;he mutations in hGH that disrupt binding to the hGH and hPRL receptors (Fig. 28). The maximal extent of 'the E~pitope for the hPRL receptor ~~t~~.'~"~4 _97_ (Fig. 25B) is approximated by mutations having less than a two-fold reduction in binding affinity. By this criteria the epitope for the hPRL receptor is essentially confined to the front face of helix 1 from F10 to Q29, i~he loop from F54 to Q68, and the hydrophilic face helix 4 from 8167 to 8178. In contrast, the hGH receptor epitope (Fig. 25A) is comprised of residues in the amino terminal region through the front face of helix 1 from I4 through M14, the loop region from F54 through S71, and the hydrophilic i:ace of helix 4 from D171 through V185.
Although further mutagenic analysis will be necessary to fill-in remaining gaps in the hPRL epitope, it is clear this epitope overlaps but is not identical to that for the hGH receptor. These data suggest that not all of the binding determinants for recognizing hGH are the same in the hGH and hPRL receptors despite them sharing 32% sequence identity in their extracellular binding domains.
Residues that. cause' large changes in receptor binding affinity may do so by indirect structural effects.
However, it is believed that most of these disruptive effects are due to local effects because all of the single mutants tested retain full binding affinity to a panel of 8 hGH monoclonal antibodies and often lead to changes in receptor preference (See Table XIX and infra) and not uniform disruptions in receptor affinity.
pes~gn of rec~.eptor specific variants of hGH.
3o A number of the single hGH mutants cause enormous changes in receptor binding preference (Table XIX).
The most notable is E174A which causes a 4-fold strengthening in affinity for the hGH receptor while weakening binding to the hPRL receptor by more than ~~~~.'~"~~4 _98-20-fold. This represents a 120-fold shift in receptor preference. Other mutations (notably R178N
and I179M) cause :hGH to preferentially bind to the hPRL receptor. Typically, the variants that cause the greatest chances in receptor specificity are located in the non-overlap regions of the two receptor epitopes.
It was reasoned that if the changes in receptor binding free energy were additive, it could be possible to <iesign highly specific variants of hGH
with only a fnw mutations. Indeed, when the two most hGH receptor selecaive single mutants (K168A and E174A) are combined, the double mutant exhibits a 1..'i 2300-fold preference for binding to the hGH receptor (Table XX). As previously indicated, the preference for binding the hPRL receptor can be enhanced by nearly 20-fold by hPL (56-64) which contains only two mutations, E..°i6D and R64M (Table XIII) . These hGH
variants (K168A,,E174A or E56D,R64M) do not substantially reduce the affinity for the preferred receptor, hG~i or hPRL, respectively. It is also possible to reduce binding to both receptors simultaneously.
Table XX.
Binding of double mutants of hGH designed to discriminate lbetween the hGH and hPRL receptors (hGHr and hPRLr). Mutants of hGH were prepared by site-directed mutagenesis, purified, and assayed for binding to the hGHr or hPRLr as described in Table XIII. Standard deviations in the determination of KD
were at or x~elow 20% of the reported value except values above 10 M which were ~ 100% of the reported value.

~(~~1.'~'~4 Table XX. (Continued) Change in receptor ~,PRLr _ hGHr preference mut ~ mut hGHr Mutant KD(~) KD(hGH) KD(~) KD(hGH) hPRLr WThGH 2.3 (1) 0.34 (1) (1) K168A, 1950 590 0.09 0.26 2300 R18N, K172A, -40,000 ~20,000 190 50 --40 For example, combining K172A, and F176A, which individually cau;~e large reductions in binding affinity to the hGH and hPRL receptors, produce much larger disruptions in affinity of 550 and !5 15,000-fold, respectively.
In all these instances the changes in the free energy of binding (ooGb.~nding) are strikingly additive (Table XXI). Additive effects of mutations have been observed in enz3~me-substrate interactions (P. J.
Carter, et al.. (1984) Cell 38, 835-840; J.A. Wells, et al., (1987) ~~roc. Natl. Acad. Sci. USA ~, 5167-5171), F~rotea:~e-protease inhibitor interactions (M. Iraskowski, et al. in Protease Inhibitors: Medical and Biological As~~ects, (1983), eds. N. Katunuma, 1!5 Japan Sci. Soc. Press, Tokyo, pp. 55-68, and protein stability (D. Shortle, et al., (1986) Proteins ~, 81-89 (1986); M.H. Hecht, J.M. Sturtevant and R.T. Sauer. Proteins 1, 43-46) and, as disclosed in these references, are most commonly found when the mutant residues function independently and are in ~I~~1.'~'~4 -loo-contact with each other. This suggests the residues paired in the multiple mutants of hGH function independently. Such additivity creates an extremely predictable situation for engineering variants of hGH
!5 with desirable receptor binding affinity and specificity.
Table XXI.
Additive effects oi: mutations in hGH upon binding to the hGH or hPRL :receptors (hGHr or hPRLr). The change in the free energy of binding (vvGbinding) for the variant relative to to wild-type hGH was calculated foom the reduction in binding affinity according to: vvGy~indin " RT In[(Kp(mut)/KD(hGH)].
The values of (KI~(mut)~KD(hGH) for the single or multiple mutant hormones were taken from Tables XIII-XX.
Change in binding free energy, vvGbinding (kcal/mol) Mutation hGH hPRLr K168A +0.04 +1.9 E174A -0.90 +1.9 K168A, E174A -0.86 +3.8 (expecaed) (act.ual) -0.80 +3.8 K172A +2.5 ~ +1.6 F176A +2.4 +1.6 K172A, F176A +4.9 +3.2 (expecaed) ( act.ual +5 . 7 +3 . 8 ) Q22N -0.06 ND

F25S +1.81 ND

D26E +0.89 ND

Q29S +0.20 ND

E33K -0.13 ND

hPRL 22-33 (expected) +2.7 -(actual) +2.6 E56A ND +0.8 R64M ND +1.8 E56A, R64M (expected) - +2.6 hPL (56-64)(actual) _ +2.0 ~~~.'~' i -lol-There are a number of other cases like hGH where two or more receptors or receptor subtypes are known to exist such ~~s fo:r the adrenergic receptors ( for review see R. J. Le~fkowitz and M. G. Caron ( 1988 ) T~.
!5 Biol. Chem. X63, 4993-4996), The IGF-I receptors (M.A. Cascieri, et al., (1989) J. Biol. Chem. ~, 2199-2202) , IL-2 receptors (R.J. Robb, et al. (1984}
J. Exp. Med. ,~, 1126-1146; R.J. Robb, et al. (1988) Proc. Natl. Acad. Sci. USA ~, 5654-5658) and ANP
1~) receptors (D. Lowe and D. Goeddel, unpublished results). In these situations it is difficult to link specific receptor function to a specific pharmacological effect. However, the use of receptor specific hormone analogs can greatly simplify this 1'.~ task. For example, catecholamine analogs were used to characterize ~-adre~nergic receptor subtypes and link receptor function to physiologic responses (for review see R.J. Lefkowitz, et al. (1983) Annu. Rev.
Biochem. 52, 159-:186). By analogy, the receptor 2n specific variants of hGH should provide a key tool for identifying other receptors for hGH, and for probing the role of the hGH and hPRL receptors in the complex pharmacology of hGH. ~ This work represents a~ sysltematic approach to identifying 2!5 receptor binding sites in hormones that permits rational design of receptor specific variants.
example 13 ~naineering Human :Prolactin to Bind to Human Growth Hormone 3n Prolactin (PRL) is a member of a large family of homologous hormones that includes growth hormones (GH), placental lactogens (PL), and proliferins.
Nicoll, C.S. et-a'.~ (1986) Endocrinol. Rev. 7, 169-203. Collectively" this group of hormones regulates 35 a vast array of physiological effects involved in ~CB~~."~"~4 growth, differentiation, electrolyte balance, and others. Chawl.a, R.K. et.al. (1983) Ann. Rev. Med. 34, 519-547: Isaksson, O.G.P. et. al. (1985) Ann. Rev.
Phvsiol. ~, 483-999. These pharmacological effects begin with binding to specific cellular receptors.
For instance, hPRL binds to the lactogenic but not somatogenic receptor and stimulates lactation but not bone growth; hGH can bind to both the lactogenic and somatogenic :receptors and stimulates both lactation and bone growth. The molecular basis for the differences in receptor binding specificity is not understood.
Cloning and Expression of hPRL.
The cDNA for hPRL was cloned from a human pituitary cDNA library in agtl0 (Huynh, T.V., et al. (1985) in DNA Cloning ech:niques: A Practical Approach, Vol. 1, D.M:. Glover, ed. (Oxford IRL Press) pp. 49-78) b;y hybridization (Maniatis, T., et al., eds. (1982) Molecular Cloning A Laboratory Manual (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY)) with ol:igonucleotide probes corresponding to 5' and 3' extremes of the published DNA sequence (Cooke, N.E., et al. (1981.) J. Biol. Chem. X56, 4007-4016).
A near full-:length cDNA clone was identified and the 720 by BstII-HindIII fragment, extending from codon 12 to 55 by past the stop codon, was subcloned into pUCil8. The sequence was determined by the dideoxy method (SangE:r, F., et al. (1977) Proc. Natl. Acad.
Sci. USA 74, 546:3-5467) and matched exactly that previously reported (Cooke, N.E., et al. (1981) Biol. Chem. x,56, 4007-4016).
The intracellular expression vector, pB0760 (Fig. 26) was created in several steps by standard methods (Maniatis, T.., et al., eds. (1982) Molecular Cloning ~~~~~.'~'~4 A Laborato ~ anual (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY) ) . The E. coli trp promoter derived from pHGH207-1 (deBoer, H.A., et al. (1982) in Promoters Structure and Function, eds. Rodriguez, °_> R.L. & Chamberli.n, M.J. (Praeger, New York) pp. 462-481) was used to transcribe the hPRL gene.
The hPRL coding ;sequence consisted of a 47 by XbaI-BstEII s~ynthet:ic DNA cassette and the 720 by BstEII-HindII:f fragment derived from the hPRL cDNA.
The synthetic DNA cassette had the sequence ***
5'-CT-AGA-ATT-ATG-TTA-CCA.-ATT-TGT-CCA-GGT-GGT-GCA-GCA-AGG-TGT-CAA
3'-T-TAA-TAC-AAT-GGT-TAA,-ACA-GGT-CCA-CCA-CGT-CGT-TCC-ACA-GTT-CAC-TG, where the initiation codon is indicated by asterisks.
1°_> The phage fl origin, pBR322 replication origin, and the pBR322 ~-lactamase gene were derived from pB0475 (Cunningham, B.C., et al. (1989) Science ~, 1330-1335).
coli cells {MM 294) containing pB0760 were grown at 37°C for 4 hr {or early log phase; A550 = 0.1 to 0.3) in 0.5 :L shake flasks containing 100 ml of M9 hycase media (Miller, J.H. (1972) ~eriments in Molecular Genet' s (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY)) plus 15 pg/ml carbenicillin.
2°> Indole acrylic acid was added (50 ~cg/ml final) to induce the trp promoter. Cells were grown an additional 6-~8 hr and harvested by centrifugation.
Cell fractionation experiments showed the hPRL was located almost exclusively in inclusion particles and represented 2-5% of the total cell protein as analyzed by S;DS-PAGE (not shown).

Purification and Refolding of hPRL. Inclusion particles containing hPRL were isolated essentially as described (Winkler, M.E., et al. (1986) Biochemistry 25, 4041-4045). Briefly, 50 g of wet cell paste was suspended in 0.25 liters, 10 mM Tris (pH 8.0), 1 mm EDTh, (TE buff er) and cells were lysed by vigorous sonication. Insoluble material was collected by centrifugation (10,000 g x 15 min) and resuspended in 25 gal of TE buffer. The suspension was layered on a 0.2 liter cushion of 50% glycerol, and centrifuged at 9,000 g x 25 min to pellet the hPRL inclus:~on particles. The hPRL from the inclusion particles (about 20% pure) was suspended in 5 ml of TE bui'fer.

The hPRL was refolded by solubilizing the inclusion particles in x.56 ml of 8N GnHCl in TE buffer plus 0.3 g reduced glut:athione (Sigma). After gentle stirring at room temperature for 30 min,, the mixture was chilled to 0°C and diluted with 844 ml of cold TE
buff er plus 0.6 g oxidized glutathione. The solution was stirred slowly overnight at 4°C, and dialyzed with 4 liters of TE buffer that was changed three times over 24 hr. Insoluble material was removed by centrifugation (10,000 x g for 20 min).
The refolded and solubilized hPRL was further purified by precipitation with (NH4)2SO4 to 45%
saturation an~3 stirred 2.5 hr at room temperature.
The precipitate was collected by centrifugation (12, 000 x g for 30 min) and redissolved in 5 ml of TE buffer. ;After 30 min at room temperature, the solution was clarified (10,000 x g for 10 min) and filtered through a millipore filter (0.45 gym) . The solution was clialyzssd against 0.5 liters of TE buffer overnight at 4°C. The hPRL (85% pure) was finally * Trade-Mark ~~~~."7"74 purified to homogeneity (>95%) by FPLC using DEAE
fast flow matrix essentially as described for purifying hGH (Cunningham, B.C., et al. (1989) Science 243, 1330-1.335).
Mutagenesis and Bindin~c Properties of hGH and hPRL
Variants.
Site-specific: mutagenesis (Zoller, M.J., et al.
(1982) Nucleic Acids Res. ~0,, 6487-6500) was carried out with the aid of a methylation repair deficient strain of ~ coli, Mut L (Kramer, B., et al. (1984) Cell 38, 879--887). Additional enrichment for mutant clones was obtained by designing mutagenic oligonucleotides to either introduce or eliminate a nearby unique restriction site so that restriction-purification or restriction-selection (Wells, J.A., et al. (1986) Phil. Trans. R. Soc.
ond. A 317, 4:15-423), respectively, could be applied to the first pool of plasmid DNA obtained after transformation of the in vitro-generated heteroduplex. All oligonucleotides were designed to have 12 by of exact match 5' to the most upstream mismatch and 10 by 3' to the most downstream mismatch. F'or mutagenesis of hGH, the previously described hGH synthetic gene contained multiple restriction taites and was cloned into the plasmid, pB0475. Variants of hGH were secreted into the periplasmic :pace of E. coli (Chang, C.N., et al.
(1987) ene ;~, 189-196) and purified as previously described.
The ICd of each analog was determined by competitive displacement of (.125I]hGH bound to the purified recombinant hGH binding protein as previously described herein and in Spencer, S.A., et al. (1988) J. Biol. Ch.em. x!63, 7862-7867. The previously described hGH binding protein (containing residues 1 to 238 of the cloned human liver receptor) was secreted and purified from E. coli as described in Fuh, G., et al. (1989) (submitted). Displacement curves were generated in triplicate and the standard deviations in the Kd values were generally at or below 20% of the raeported values and did not exceed 50% of the reported value except when Kd values were greater than 7.0 ~cM.
The concentrations of hPRL and hPRL mutants were determined by A2g0 using a calculated extinction coefficient of ~S(0.1%,280) - 0.9 (Wetlaufer, D.B.
(1962) Adv. in Prot. Chem. 17, 303-390). This was adjusted accordingly when variants contained mutations in aromatic residues. Concentration values determined by absorbance agreed to within 10% to those determi:ned by laser densitometry of proteins run on SDS-PAnE and stained with Coomassie blue for hGH. Circula~~ dichroic spectra were collected on an Aviv Cary 60 s:pectropolarimeter.
In order to probe which of the divergent residues in hPRL were moat dissruptive for binding to the hGH
receptor (Fig. 27), a number of hPRL residues were first introduced into hGH (Table XXII) .
*Trade-Mark ~~.$'~~..'a "i 4 -l07-Table XXII.
Comparison of h1?RL and alanine substitutions introduced into hGH:
hGH variant Kd(nM) ~ mut ICd (hGH) WT 0.34 (1) I58L 0.58 1.7 I58A 5.6 16 R64K 0.20 0.6 R64A 7.1 21 F176Y 2.9 8.6 F176A 5.4 16 R178K 1.7 5.1 R178N 2.9 8.5 Whereas single a7Lanine substitutions in hGH at positions 58, 64, 176 and 178 strongly disrupted receptor binding: substitutions of hPRL residues into hGH at these positions had less of an effect. The !5 largest effecas for hPRL substitutions were in the helix 4 residues that included positions 176 and 178.
These data suggested that residues in the helix 4 region of hPRL could best account for the lack of binding to the hGH receptor.
lid The recombinant hPRL retained native-like structural and functional properties. First, the near and far ultraviolet c:D spectra (Fig. 28) are identical to published spectra of natural hPRL (Bewley, T.A.
( 1979 ) in ~ec:ent rogress in Hormone Research, vol .
1!5 35, pp. 155-213, Acad. Press, N.Y.). The far ultraviolet spectra is similar to hGH, suggesting a similar 4-helix bwldle structure, although important differences in the mean residue ellipticity at 208 and 224 nm have been noted (~.). These hormones 2~~ differ markedly in the near ultraviolet CD which reflects variation in number and microenvironment of the aromatic residues between hGH and hPRL. In other ~~~."~"~4 -l08-studies (not shown), the recombinant hPRL retained full immunological cross-reactivity in an hPRL ELISA, and was equipotent with hGH in causing rat lymphoma Nb2 cells to proliferate (Tanaka, T., et al. (1980) J. Clin. Endo. Metab. ,5~, 1058-1063). Upon reduction, t:he purified hPRL showed a pronounced retardation in mobility by SDS-PAGE (as seen for hGH) suggesting that disulfide bonds had formed (Pollitt, S., et al. (;1983) J. Bacteriol. ,~, 27-32) . Amino terminal sequence analysis showed that the intracellularly expressed hPRL retained the amino terminal methioninEe: however, as with methionyl-hGH
(Olson, K.C., et al. (1981) ature (London) 293, 408-411), this does not apparently affect its structure or function.
Binding of hPRL to the hGH binding protein is reduced by more than 105-fold compared to hGH
(Table XXIII) which is below the detection limit of our binding assay.

~~~~.~~I ~:

Table XXIII.
Engineering residues in hPRL to permit binding to the hGH binding proteinl hPRL Variant Kd (nM)2 Kd~mut) Kd(hGH) hPRL >40,000 >100,000 WT

A H171DN175TY176F 4,900 14,000 =

B A + K1?8R 220 660 =

B hGH (184-188) 260 740 +

hGH(54-74) -25,000 -66,000 B hGH (54-74) 2,000 5,800 +

B H54FS56E:L58I: 36 110 +

E62S:D63N:Q66E

B H54F:S56E:L58I 670 2,000 +

C B + E174A 68 200 =

D C + E62S:D63N:Q~66E 6.2 = 2.1 D H54F 4.4 13 +

D S56E 2.5 7.4 +

D L58I 3.6 11 +

D A59P 2.5 7.4 +

D N71S 3.6 11 +

D L179I 2.1 6.2 +

lMutants of h,PRL were generated, purified and analyzed as described. Multiple mutants are indicated by a series of single mutants (Table XXII ) separated by colons. Codon numbering is based upon the hGH sequence (fig. 2).
2Average standard errors are at or below 20% of the reported values, except in cases where the Kd exceeds 1 ~M where it can be as large as 50%, and errors are much larger still when Kd exceeds 10 ~M.

'~~~~.'~"~4 -l o-A combination of three divergent residues in helix 4 from hGH (H~.71D, N175T, and Y176F) were introduced into hPRl. Alani.ne scanning mutagenesis and hPRL
substitution~> (Table XXII) had shown that these residues were very important for binding hGH to the hGH receptor., This triple mutant of hPRL exhibited detectable binding to the hGH binding protein albeit 14,000-fold weaker than hGH. Installation of another important hs~lix 4 residue (K178R) to produce a tetramutant (called variant B in Table XIII) further strengthened binding to a level now only 660-fold below wild-type hGH. Additional incorporation of hGH
residues 184 to :L88 into hPRL variant B did not enhance binding to the hGH binding protein. However, introduction. of E174A to give hPRL variant C
(Table XXIII) caused an additional 3.5-fold increase in binding affinity to the hGH binding protein as was found when E;174A was incorporated into hGH.
Having engin~eered~binding with the helix 4 region, the loop region containing residues 54 to 74 was analysed. Complete replacement of the loop region in hPRL with tlhe ser,~uence from hGH (hGH ( 54-74 ) in Table XIII) cave barely detectable binding to the hGH
binding protein. When this mutant was combined with variant B, the binding affinity increased substantially. However, this new variant [B plus hGH
(54-74)] was reduced in binding affinity by almost 10-fold from variant 8 alone. Thus, it appeared that some of the hGH rEaidues in the 54-74 loop were not compatible with the hGH substitutions in helix 4. We then selected from the 54 to 74 loop of hGH only those seven residues that were shown by alanine scanning mu.tagenesis to most greatly influence binding. Although the R64A mutation in hGH caused more than a 20-fold reduction in binding affinity, ~~U1'~'~4 the R64K variant_ of hGH (which is an hPRL
substitution) slightly enhanced binding to the hGH
binding protEain (Table XXII) . The Lys64 in hPRL
therefore was left unchanged. As a consequence, only !i six of the seven substitutions from hGH were incorporated into hPRL that were most disruptive when changed to al.anine in hGH. This new mutant (B plus H65F:S56E:L58I:E5EiS:D68N:Q66E) binds fifty-fold stronger than B plus hGH (54-74) and was only 110-fold reduced in binding affinity from wild-type hGH
(Table XXIII). However, this represented only a modest improvement (six-fold) over variant B alone which was less than expected for strongly favorable interactions previously observed in the loop region 1!i for hGH. Therefore, the six mutations within the loop were further dissected and revealed that the combination c>f H54F:S56E:L58I plus variant B bound three-fold weaker than variant B alone. Finally, incorporating the: mutations E62S:D63N:Q66E into 2c) variant C (to give variant D) produced an analog with highest affinity that was only 6-fold reduced in binding affinity relative to hGH. Additional single mutations (H_°°.4F, S56E, L58I, A59P, N71S and L179I) did not enhance the binding affinity of hPRL variant 2!i D to the hGH binding protein. The conformation of variant D was virtually indistinguishable from native hPRL by CD spectral analysis (Fig. 28) or by ELISA
reactivity (not shown).
These stud ies demonstrate the feasibility of 30 recruiting binding properties fox distantly related homologs using only functional information derived from site-directed mutagenesis experiments. Alanine scanning mutagenes.is of hGH provided a systematic analysis of side-chains that were important for 3!5 modulating binding of hGH to its receptor (Fig. 27).

i ~~~.~'~~

This information highlighted a number of residues in hPRL that could acc;ount for its inability to bind to the hGH receptor (Fig. 29). However, further analysis showtad that the alanine substitutions in hGH
were more disruptive than the hPRL substitutions in hGH (Table XXII). Furthermore, some of the hPRL
substitutions were considerably more disruptive than others for binding .affinity, especially when a larger side-chain was present in hPRL. For example, the 1C~ conservative (but larger) F176Y mutation in hGH
caused an eight-fold reduction in binding affinity with the hGfi receptor, whereas the smaller R64K
substitution showed slightly enhanced binding affinity. Thus, the analysis of disruptive hPRL
1°_~ substitutions in hGH suggested the introduction of the cluster of divergent residues in helix 4 to initially achieve binding affinity for hPRL. This was very important because no binding to the hGH
receptor with wild--type hPRL had been observed, and 2C~ it was necessary to introduce several hGH
substitutions simultaneously into hPRL in order to bring the binding affinity within the range of the assay used (Kd s50 ,uM) .
Readily detectable binding affinity was engineered 2_°°. into hPRL by incorporating functionally important residues into helix 4. However, engineering the loop region between 54-7~4 turned out to be more difficult.
Installing tlhe entire loop from hGH into hPRL
produced less enhancement in binding than expected, 30 and was disruptive to binding when combined with the optimized helix 4 variant B. Our data suggest that the 54-74 loop structure in hPRL is supported by other interactions in the protein. This problem was solved in stages. First, only those six loop 3_°°. residues from hGH that the alanine scan together with ~~~~,'~'~4 the hPRL substitutions in hGH had identified to be important were introduced into hPRL. Although this improved the situation, the combination of some of these hGH mutation; (narrowed down to H54F, S56E, and L58I) were disruptive to hPRL. These data suggest that some of the reaidues in the loop are crucial for its structure and are better off being left alone.
A number of iterative cycles of mutagenesis were necessary to converge upon a combination of residues that permitted tight binding of hPRL to the hGH
receptor. This strategy relies on the assumption that the mutational. effects will be somewhat additive as was, in fact, observed. For example, E174A
mutation enhanced t:he binding three to five-fold when added to either hPF;L variant C or hGH. Moreover, the product of tree disruptive effects of the H54F, S56E, and L58I single mutants to variant D (4.4-fold) is about the same as the disruption caused by the combination of all three mutations added to variant B
(3-fold).
Even though variant D is only six-fold reduced in binding affinity, there are several other residues that could bc: incorporated into variant D to try to improve further on the binding, such as V14M and H185V: these are sites where alanine substitutions in hGH cause two to five-fold reductions in binding of hGH (Fig. 29). Although a high resolution structure would have aided in the design process, it was clearly not Easent:ial. The cumulative nature of the 3o mutational effects allows one to converge upon the binding property i.n much the same way as proteins evolve, by cycles of natural variation and selection.

~~~~.'~"~4 m -114-Previous protein engineering experiments have shown it is possible using high resolution structural analysis to virt:ually exchange the substrate specificity of natural variant enzymes by !5 site-directed mut:agenesis of substrate contact residues (We:lls, ~T.A., et al. (1987) Proc. Natl.
i'~cad. Sci. USA $~" 5167-5171; Wilks, H.M., et al.
(1988) Science ~,, 1541-1544). Similarly, others have shown that binding properties can be engineered by replacement of entire units of secondary structure units including antigen binding loops (Jones, P.T., et al . ( 1586) Nature 321, 522-525) or DNA
recognition helices (Wharton, R.P., et al. (1985) Nature 316,601-605). However, to recruit the hGH
1!p receptor binding properties into hPRL required selective residue replacements within the structural scaffold of hPRL. Furthermore, the CD spectral data show that the overall structure of the hPRL variant D
resembles more clo:>ely the structure of hPRL not hGH
even though it attains binding properties like hGH.
The fact that the binding specificity for the hGH
receptor could be incorporated into hPRL confirms the functional importance of particular residues for somatogenic ~__~eceptor binding. These studies also 2!5 provide compelling proof for structural relatedness between hGH and hPRL despite them having only 23%
identity. 'this provides a rational approach to access new receptor binding functions contained within this hormone family starting with either a growth hormone, prolactin, proliferin or placental lactogen scaffold. Such hybrid molecules should be useful for distinguishing receptor binding and activation as well as the pharmacological importance of receptor ;subtypes. These analogs could lead to ~~~~~.'~"~4 the design of new receptor-specific hormones having more useful properties as agonists or antagonists.
Example 14 Recruitment of binding properties of human growth hormone into human vplacental lactoaen.
Human placental lactogen (hPL) is reduced over thirty-fold in binding affinity compared to hGH for the hGH recepitor (G. Baumann, et al., (1986) J. Clin.
Endocrinol. Metab. ~,~, 134: A.C. Herington, et al.
1C~ (1986) J. Clin. Invest. 77, 1817). Previous mutagenic studies showed the binding site on hGH for the hGH receptor is located primarily in two regions (including reaidue:a 54-74) and 171-185) with some minor determinants near the amino terminus (residues 1°_. 4-14 ) .
The overall sequence of hPL is 85% identical to hGH.
Within the three regions that broadly constitute the receptor binding ep~itope on hGH, hPL differs at only seven positions and contains the following 2C> substitutions: P2Q, I4V, N12H, R16Q, E56D, R64M, and I179M. (In this nomenclature the residues for wild-type hGH is given i.n single letter code, followed by its position :in mature hGH and then the residue found in hPL; a similar nomenclature is used to describe 2°_. mutants of hGH). Single alanine substitutions have been producEad in hGH at each of these seven positions. Of these, four of the alanine substitutions were found to cause two-fold or greater reductions in binding affinity including I4A, E56A, 3C~ R64A, and I179A. Generally, the alanine substitutions have a greater effect on binding than homologous substitutions from human prolactin.
Therefore, th.e effect of some of the substitutions from hPL initroduced into hGH were investigated.

~,.~ - .. ir~ a W 'r' ~\.a .

Whereas the I179A substitution caused a 2.7-fold reduction in affinity the I179M caused only a slight 1.7-fold effect. However, the R64A and R64M
substitutions caused identical and much larger !5 reductions (about 20-fold) in binding affinity.
Moreover, the: double mutant (E56D:R64M) in hGH was even further reduced in affinity by a total of 30-fold (Tab7.e I). Thus, E56D and R64M primarily determine t;he differences in receptor binding affinity between hGH and hPL. The double mutant D56E, M64R i:n hPL therefore substantially enhances its binding affinity for the hGH receptor.
Additional modifications such as M179I and V4I also enhance binding of hPL to the hGH receptor.
1!i Example 15 effect of amino acid replacement at position 174 on binding' to the human crrowth hormone.
As previously indicated, replacement of G1u174 with Ala(E174A) resulted in more than a 4-fold increase in 2t) the affinity of human growth hormone (hGH) for its receptor. 'To determine the optimal replacement residue at position 174 hGH variants substituted with twelve other residues were made and measured to determine their affinities with the hGH binding 2!5 protein (Table XXIV). Side-chain size, not charge, is the major' factor determining binding affinity.
Alanine is the optimal replacement followed by Ser, Gly, Gln, Asn, Glu, His, Lys, Leu, and Tyr.

r~

Table XXIV.
S_~3e chain Mutants Charge Size(A3)b Kd(nM)c Kd(wi dutype) E174G 0 0 0.15 0.43 E174A 0 26 0.075 0.22 E174S 0 33 0.11 0.30 E174N 0 69 0.26 0.70 E174V 0 76 0.28 0.80 wild-type - 89 0.37 1.0 E174Q 0 95 0.21 0.60 E174H 0 101 0.43 1.2 E174L 0 102 2.36 6.4 E174K + 105 1.14 3.1 E174R + 136 NE -E174Y 0 137 2.9 8.6 a Mutations were generated by site-directed mutagenesis (C:arter, P., et al. (1986) Nucleic Acid Res. 13, 4431-4443) on a variant of the hGH
gene that contains a KpnI site at position 178 cloned into p~B0475. Oligonucleotides used for mutagenesis had the sequence:
5'-AC-AA~s-CTC-NNN-ACA-TTC-CTG-CGC-ATC-GTG-CAG-T-3' where NNN represents the new codon at position 174 and asterisks indicate the mismatches to eliminate the KpnI site starting at codon 178.
Mutant codons were as follows: Gln, CAG; Asn, AAC: Ser, AGC; Lys, AAA: Arg, AGG: His, CAC:
Gly, GGG: Val, GTG; Leu, CTG. Following heteroduplex synthesis the plasmid pool was enriched for the mutation by restriction with KpnI to reduce the background of wild-type sequence. A1:1 mutant sequences were confirmed by dideoxy segvence analysis (Sanger, F., et al.
(1977) g oc. atl. Acad. Sci. USA 74, 5463-5467.
b Side-chain packing values are from C. Chothia (1984) ~nnu. .ev. Biochem. ~3_, 537.
c Dissociation constants were measured by competitive diplacement of [1251]hGH from the hGH binding protein as previously described. NE
indicates that: the mutant hormone was expressed at levels too low to be isolated and assayed.

~;~2'~'~4 -lls-Example 16 The hGH variants shown in Table XXV were constructed.
Their relativity potency as compared to wt-hGH are shown.
Table XV.
Relative potency in hGH mutant rat weight gain assay F97A 0.87 S100A 2.12 L101A 3.03 V102A 1.39 Y103A 1.73 T175S 1.21 Having described 'the preferred embodiments of the present invention, it will appear to those ordinarily skilled in tile art that various modifications may be made to the disclosed embodiments, and that such modifications. are intended to be within the scope of the present invention.

Claims (29)

1. A method for identifying at least a first unknown active domain in the amino acid sequence of a parent polypeptide, said active domain interacting with a first target, said method comprising;
a) substituting a first selected amino acid segment of said parent polypeptide with a first analogous polypeptide segment from an analog to said parent polypeptide to form a first segment-substituted polypeptide, said parent polypeptide and said analog having a different interaction with said first target;
b) contacting said first segment-substituted polypeptide with said first target to determine the interaction, if any, between said first target and said segment-substituted polypeptide;
c) repeating steps a) and b) using a second analogous polypeptide segment from an analog to said parent polypeptide to form at least a second segment-substituted polypeptide containing a different analogous polypeptide segment than said first segment-substituted polypeptide; and d) comparing the difference, if any, between the activity relative to said first target of said parent polypeptide and said first and second segment-substituted polypeptides as an indication of the location of said first active domain in said parent polypeptide.

2. The method of Claim 1 wherein said unknown active domain comprises at least two discontinuous amino acid segments in the primary amino acid sequence of said parent polypeptide.

3. The method of Claim 1 wherein at least said first selected amino acid segment of said parent polypeptide contains at least one amino acid residue located on the surface of a native-folded form of said parent polypeptide.

4. The method of Claim 3 further comprising repeating steps a) and b) until substantially all of the amino acid residues on said surface of said parent polypeptide have been substituted by said analogous amino acid segments.

5. The method of Claim 1 further comprising repeating steps a) and b) until about 15-100% of the amino acid sequence of said parent polypeptide has been substituted by said analogous amino acid segments to form the first segment-substituted polypeptide.

6. The method of Claim 1 further comprising repeating steps a) and b) until about 60-100% of the amino acid sequence of said parent polypeptide has been substituted by said analogous amino acid segments to form the first segment-substituted polypeptide.

7. The method of Claim 1 further comprising identifying a second unknown active domain of said parent polypeptide, said second active domain interacting with a second target, said method comprising repeating steps a) through d) with said second target.

8. The method of Claim 1 further comprising identifying at least a first active amino acid residue within said first active domain, said method comprising:
e) substituting a scanning amino acid for a different first amino acid residue within said first active domain to form a first residue-substituted polypeptide;
f) contacting said first residue-substituted polypeptide with said first target to determine the interaction, if any, between said target and said residue-substituted polypeptide;
g) repeating steps e) and f) to substitute a scanning amino acid for at least a second amino acid residue within said first active domain to form at least a second residue-substituted polypeptide; and h) comparing the difference, if any, between the activity relative to said first target of the parent polypeptide and each of said first and second residue-substituted polypeptide as an indication of the location of the active amino acid residue in said first active domain.

9. The method of Claim 8 further comprising repeating steps a) through h) with a second target substance to identify a second active domain and at least one active amino acid residue within said second active domain.

10. The method of Claim 9 further comprising the step of substituting at least one of said active amino acid residues in said first active domain with a different amino acid to produce a polypeptide variant having a modified interaction with said first target but which retains substantially all of the interaction of said parent polypeptide with said second target.

11. The method of Claim 10 further comprising the step of substituting at least one of said active amino acid residues in said second active domain with a different amino acid to produce a polypeptide variant having a modified interaction with said first and said second target.

12. The method of Claim 9 wherein said first and said second active domains have at least one common active amino acid residue, said method further comprising substituting at least said one common active amino acid residue with a different amino acid to produce a polypeptide variant having modified interactions with each of said first and said second targets.

13. The method of Claim 9 wherein said first and said second active domains have at least one common active amino acid residue, said method further comprising substituting at least one amino acid residue in said first active domain not comprising said at least one common active amino acid residue with a different amino acid to produce a polypeptide variant having a modified interaction with said first target.

14. A method for identifying at least one active amino acid residue in a parent polypeptide, said method comprising:
(a) substituting a scanning amino acid for a first amino acid residue at residue number N within said parent polypeptide to form an N-substituted polypeptide;
(b) substituting a scanning amino acid for each of the amino acid residues at residue numbers N+1 and N-1 to said first residue to form respectively N+1- and N-1-substituted polypeptides;
(c) contacting each of said substituted polypeptides with a target to determine the interaction, if any, between said target and said substituted polypeptides;
(d) comparing the difference, if any, between the activity of the parent polypeptide and said substituted polypeptides with said target;
(e) repeating steps (b) through (d) for increasing residue numbers if said activity difference between said target and said N+1 substituted polypeptide is greater than two-fold and for decreasing residue numbers if said activity difference between said target and said N-1 substituted polypeptide is greater than two-fold.

15. The method of Claim 14 wherein steps (b) through (d) are repeated until at least four substituted polypeptides containing the substitution of a scanning amino acid at four consecutive residues are identified having less than a two-fold activity difference as compared to said parent polypeptide.

16. The method of Claim 1, 8 or 14 wherein said parent polypeptide is selected from the group consisting of human growth hormone, human prolactin, .alpha.-interferon, .gamma.-interferon, tissue plasminogen activator, IGF-1, EGF, CD-4, TNF, GMCSF, TGF, follicle stimulating hormone, luteinizing hormone, atrial natriuretic peptide and placental lactogen.

17. The method of Claim 16 wherein parent polypeptide is human growth hormone, human placental lactogen or human prolactin.

18. The method of Claim 1 wherein said parent polypeptide is human growth hormone and said analog is selected from the group consisting of human placental lactogen, porcine growth hormone and human prolactin.

19. The method of Claim 8 or 14 wherein said scanning amino acid is an isosteric amino acid.

20. The method of Claim 8 or 14 wherein said scanning amino acid is a neutral amino acid.

21. The method of Claim 20 wherein said neutral amino acid is selected from the group consisting of alanine, serine, glycine and cysteine.

22. The method of Claim 21 wherein said scanning amino acid is alanine.

23. The method of Claim 1, 8 or 14 wherein said activity is measured in an in vitro or in vivo assay.

24. The method of Claim 23 wherein said parent polypeptide is a hormone and said activity is measured in an in vitro assay using a soluble hormone receptor.

25. The method of Claim 24 wherein said hormone is human growth hormone and said soluble hormone receptor is shGHr.

26. The method of Claim 24 wherein said hormone is human growth hormone and said soluble hormone receptor is shPRLr.

27. The method of Claim 1, 8 or 14 wherein said activity indicates the binding of said target to said parent polypeptide or the catalysis of said target by said parent polypeptide.

28. The method of Claim 27 wherein the activity between said target and said substituted polypeptide is increased as compared to said parent polypeptide.

29. The method of Claim 27 wherein the activity between said target and said substituted polypeptide is decreased as compared to said parent polypeptide.